The latest funding proposals fall far short of what many contend is needed to prepare for the decennial count. In February, the Government Accountability Office added the program to its “high risk” list. And just last week, Director John Thompson surprised Census observers when he announced that he would resign at the end of next month.
Former agency officials and Census advocates are worried that inadequate preparation could potentially spell significant problems for the accuracy of the count. Given that congressional redistricting and funding for hundreds of federal programs all rely on the decennial Census, the reliability of the numbers carry far-reaching consequences.
Certain segments of the country, in particular, are much more vulnerable to inaccuracies or underreporting than others. Racial and ethnic minorities, poorer households and those on the margins of society could be disproportionately hurt if the Census reaches fewer households.
“The lack of resources will weaken the operations specifically designed to make the Census more accurate in historically undercounted communities,” says Terri Ann Lowenthal, a former staff director of the House subcommittee tasked with Census oversight.
Already, the agency has terminated field test operations planned for this year in Puerto Rico, North Dakota, South Dakota and Washington state.
The federal government has historically funded the Census Bureau in cycles, allocating large increases in the years leading up to the decennial count to support preparation activities. But the Trump administration’s budget proposal for next fiscal year only allocates $1.5 billion, about the same as the current budget. By comparison, the agency's budget nearly doubled between fiscal years 1997 and 1998, and increased more than 60 percent in 2008 in the lead up to the 2010 Census, according to Office of Management and Budget records.
Kenneth Prewitt, who headed the agency until 2001, says underfunding the Census could hinder its reliability in several ways.
“It would inevitably exaggerate the problems that the Census will always have to some degree,” he says. “It can't be fair if it has a serious overcount or undercount.”
Disadvantaged communities, such as low-income households, renters, those without phones and non-English speakers, generally have lower response rates. The single largest expense for the Census is the small army of enumerators temporarily hired to canvas neighborhoods, so cutbacks could particularly affect counts in these communities.
Many of the Census Bureau’s outreach efforts aimed at ensuring all people are counted, such as advertising and language assistance, are further vulnerable to funding cuts.
“Right now, they don’t have enough money to do the advertising,” says Prewitt, now a Columbia University professor. “When we had the money to run ads [in five languages], it made a huge difference.”
Census experts say without such outreach efforts or the ability to send enough enumerators to tally nonresponding households, discrepancies in undercounting of disadvanted communities could worsen.
According to the Census Coverage Measurement survey following the last decennial Census, the national net undercount of the black population was 2.1 percent. Similarly, Hispanics -- one of the fastest-growing demographics -- were undercounted at a rate of 1.5 percent in 2010.
| Demographic Group | 1990 Net Undercount | 2000 Net Undercount | 2010 Net Undercount |
|---|---|---|---|
| White (non-Hispanic) | 0.68% * | -1.13% * | -0.84% * |
| Black (non-Hispanic) | 4.57% * | 1.84% * | 2.07% * |
| Asian (non-Hispanic) | 2.36% * | -0.75% | 0.08% |
| American Indian on Reservation | 12.22% * | -0.88% | 4.88% * |
| Native Hawaiian or Pacific Islander | 2.36% * | 2.12% | 1.34% |
| Hispanic | 4.99% * | 0.71% | 1.54% * |
SOURCE: U.S. Census Bureau
Disparities are particularly severe for more narrow demographic groups. Black men between the ages of 30 and 46 were undercounted at a rate of 10 percent, while 5.9 percent of those in their late teens and 20s weren’t counted. Hispanic men of the same ages are also characterized by high rates of undercounting exceeding 5 percent.
At the same time, non-Hispanic whites are slightly overcounted. That’s typically a result of individuals owning multiple homes and filling out forms for both addresses. White women over age 50 are most often counted more than once, with an estimated net overcount of 2.3 percent in the last Census.
The gaps between the overcount and undercount represent the extent to which populations are underrepresented.
At the national level, undercounts and mistakes made on Census forms resulting in overcounting effectively cancel each other out to some degree. But they’re much more problematic at the local level, Lowenthal says, where populations tend to be more homogeneous and may be subject to widespread underreporting.
“One can look at the patterns and understand that there clearly are consequences for communities,” she says.
Any significant undercounting could carry long-term political implications since congressional redistricting is tied to the decennial Census. Many groups historically subject to undercounting tend to vote Democratic. But those in rural areas, where enumerators also face challenges in contacting households, are more conservative.
Typically, statistical sampling would be used to adjust for the undercount. Prior to the 2000 Census, the Commerce Department and Clinton administration backed plans to utilize sampling methods. But the House of Representatives, led by Speaker Newt Gingrich, sued to block it and later threatened a government shutdown. The Supreme Court subsequently ruled the agency must rely on a traditional head count, without an adjustment, for apportionment of House seats.
In addition to political implications, Census counts further determine how a large portion of federal dollars is spent.
A Brookings Institution report estimated 215 federal programs utilized Census data to appropriate federal assistance totaling $446.7 billion in fiscal year 2008. Medicaid accounts for the single largest program. Highway construction assistance, special education grants and the Children's Health Insurance Program are among other major programs relying on the census.
Population counts for different racial groups are also critical in enforcing civil rights laws, such as the Voting Rights Act.
On numerous instances in recent years, Congress has directed the agency not to spend more on the 2020 Census than it did for the 2010 count. But that doesn’t account for inflation, and projections further indicate there will be about 8 percent more Americans to count.
The agency plans to implement several new cost-savings approaches for the 2020 Census, such as allowing for responses to be submitted via phone or the internet. Officials also plan to utilize administrative and third-party records to aid in identifying vacant housing units and occupied units that don’t initially respond.
Census experts emphasize that testing is especially crucial this time around given these changes and the newer technologies being used for the first time.
“We were still doing the Census the old-fashioned way,” says Steve Murdock, who served as director of the Census Bureau from 2007 through 2009. “Now we’re doing it in ways that really require some additional and careful testing.”
Murdock says the agency had planned to incorporate a much more computerized process prior to the 2010 Census. But testing revealed problems, such as not being able to distinguish legitimate responses from fake ones.
Next year, the agency plans to test its methods in three jurisdictions for its “end-to-end” test, a trial run that’s done two years before every decennial Census.
“Everyone wants an accurate Census,” Murdock says. “The question is, at what cost is it worth?”
