What if the next U.S. census was canceled because it was too expensive? As farfetched as that may sound to demographers and social scientists, a congressional panel heard yesterday that such a dire scenario is possible unless the Census Bureau manages to significantly lower the soaring costs of doing a high-quality nose count.
The 2010 census clocked in at $13 billion, noted Senator Tom Carper (D-DE), chair of the government information panel for the Committee on Homeland Security and Governmental Affairs, during a hearing on the lessons learned from the latest decennial count. "I've been told that the 2020 Census could cost as much as $30 billion. That is not acceptable," he said. Legislators were equally unhappy with data from a new Government Accountability Office report that showed that the average cost per household for last year's census was $98, up sharply from $70 in 2000 and more than six times the $16-per-household cost in 1970.
Several witnesses told the panel that the bureau needs to move into the 21st century in improving its approach to finding residents and offering them multiple ways to answer the questionnaire.
Census officials also need to make those decisions early enough in the decadal cycle, they said, to ensure that the new approaches have been sufficiently road-tested before the actual deployment of the next census in April 2020.
Recommendations include encouraging residents to reply initially via the Internet, making use of administrative records now held by other federal agencies to flesh out those responses, and automating its field operations. The last is where the dollars really add up: It costs an average of $57 to knock on the door of each person who has not mailed in the 10-question form. The bureau was also urged to use its monthly American Community Survey (ACS), which replaced the long form of the census, to try out some of these new ideas.
"The Census Bureau needs to put some stakes in the ground that should not be the subject of debate once agreed upon," urged Thomas Cook, who led a National Academies' report released late last month titled Change and the 2020 Census: Not Whether But How. Referring to next month's Canadian census and a census taken last year in Brazil, Cook noted that the bureau "should not reinvent the wheel, but build on and work from external experience."
Census Director Robert Groves says he couldn't agree more that major changes are needed to make the 2020 census viable. "It's not likely we'll get more resources, so we must become more efficient," he told the panel. But accomplishing that goal won't be easy. Speaking with ScienceInsider after the hearing, Groves outlined the challenges facing the bureau and how he hopes to address them"
How much cheaper do we need to make it? You'll have to ask Congress. But I do know a few things. I know that the cost of every U.S. sample survey is going up. Why is that? Have we forgotten how to do surveys? No. It's because contacting people is getting tougher every year, so all of our survey costs are zooming up. It's not because we're stupid. If we do nothing, the cost will go up.
So stopping that increase is a wonderful goal. But how much, we don't know. We're going to simulate cutting out some of the costly operations, and see how much we save. And then we'll ask the American public if this is an acceptable change.
The current drive to shrink federal spending promises to complicate Groves's job. Just this week, for example, the House republicans proposed trimming $22 million from the Census Bureau's current $964 million operations budget for surveys as part of its plan to reduce the 2011 budget. And Groves's request for $67 million in the 2012 budget for his agency to begin using the ACS as a test bed is also likely to be a target for congressional budget-cutters who regard the 2020 census as something far in the future. "We think we can reduce these costs," says Groves, "but we need an ongoing dialogue with Congress, especially in the early stages of the decade, when they tend not to pay attention."