The decennial census is one of the federal government’s most challenging operational undertakings. The census ramps up operations over a decade to canvass addresses, test questions and procedures, hire hundreds of thousands of workers, build partnerships, open field offices, implement new methodological and technological innovations, enumerate the country, and provide final counts to the federal government—all to advance the constitutionally mandated mission to count all people in the country on April 1 of every year ending with a zero.

It is an extraordinary feat that added over half a million temporary workers (PDF) to census payrolls at its peak in 2010, producing a bump in the nation’s overall labor force participation rate.

Given how complicated the logistics are to produce as accurate a count as possible, funding is critical, and a predictable funding stream is necessary for the planning and execution of a successful census. The 2020 Census has lacked this from the start. As Arloc Sherman at the Center on Budget and Policy Priorities has shown, planning and final testing are critical in the latter half of the decade (i.e., years 6–9). But in this past decade, funding for the run-up to the 2020 Census fell below levels provided in the last three decennial census cycles.

We updated the Center on Budget and Policy Priorities analysis to look at the entire decade of US Census Bureau funding relative to the last two censuses (2000 and 2010). We looked at the run-up years—years 6 to 9—compared with the year 1 to year 5 average, with the census itself in year 10.

Pre-2020 spending in years 6 and 7 was consistent with 2010 and ahead of 2000. But in years 8 and 9, spending did not ramp up as much as previous censuses had. Years 8 and 9 were, and continue to be, a time to test new cost-saving measures. With the slower lead-up in year 8 and 9 budgets, there was less money for testing, so we know less about how well these cost-saving innovations will work.

If we used the 2010 Census as a baseline, we would know about how much money (adjusted for inflation) would be needed to run the upcoming census. But technologies and methodologies never stay static when operating the decennial survey; they change to keep pace with innovations over the decade, potentially improving the count. What remains unclear is if the projected cost savings for innovations introduced to the 2020 Census balance out any reductions in the budget because these innovations have not been tested at scale.

For example, people might be less willing than in the past to respond to mailed census forms but more responsive to an internet-based form. The internet self-response option should cut down on the cost (PDF) of mailing forms, in-person visits, and scanning and entering data—processes critical to previous censuses. But because it has not been previously implemented at this level, and with response rates slipping year after year, we don’t know how much money will be saved in the end.

Similarly, the use of government administrative records could reduce expenses related to address verification and nonresponse follow-up. But again, it remains untested in a decennial census environment, so the benefits are difficult to estimate.

Because of budget cuts, census officials have had to scale back key components of testing and outreach. The 2018 end-to-end tests (PDF), which act as dress rehearsals for the decennial enumeration, were originally slated for three areas: Bluefield-Beckley-Oak Hill, West Virginia; Pierce County, Washington; and Providence County, Rhode Island. In the end, only Providence received the end-to-end test.

This shortchange raised concerns because it was a missed opportunity to confirm data collection methods and to note public response to new outreach strategies in rural and tribal areas. The Bureau decreased the number of regional and area offices for the 2020 Census, and it is uncertain if the potential benefits of increased automation compensate for a reduction in local brick-and-mortar presence.

Census funding needs to be predictable and forward-facing enough so that census experts can test new methodologies, question wording changes, and streamline technological innovations that could make for a more modern and cost-effective census. And if the Census Bureau does not get this money throughout the decennial process, the long-range potential benefits will not be realized. It is encouraging that Congress and the administration have slated additional funds for the 2020 Census in the coming year, but it could be too little too late.