The Internet Age is upon us. But rather than circulating online, the 23rd Decennial Census stuck with the tried-and-true, and flooded the U.S. Postal Service March 16 through 18 with surveys en route to more than 120 million households nationwide. The 10-question form, which probes for demographic information such as age, sex and race, will help determine how more than $400 billion will be allocated to communities across the country. Citizens and noncitizens alike are required by law to complete the form and mail it back to the U.S. Census Bureau in the accompanying prepaid envelope. That's a lot of mail, but that's not all of it.

In case the mail at your household gets picked up and thrown into the "we'll get to it later" pile, the Census Bureau took the extra step this year of sending out a "heads-up" letter in advance—a "state-of-the-art practice in survey research," according to Census Director Robert Grove's blog—to encourage participation. And in case that's not enough paper for you, an extra nudge was mailed out the week of March 22. This might sound excessive, but the mail-out/mail-back response rate for the 2000 census was only 65 percent, and the missing data has to be collected in person by enumerators at a cost of about $57 per household. So the nudge "more than pays for itself," Groves says.

Mandated by the U.S. Constitution and developed in 1790 as a cornerstone of representative democracy, the census was initially carried out by trained enumerators taking a physical head count across the country. The move to the mail-out/mail-back survey in 1970 was driven by cost. But the 2010 survey’s $14.5-billion price tag has the Bureau considering a next-gen cost-effective replacement. Three in four people living in the U.S. have Internet access, according to Nielsen NetRatings. Going online and paper free could cut the printing and mailing costs of $75 million and $59 million, respectively.

Alternative data collection methods, including digital ones, will be a major topic of discussion at a 2020 Decennial Census advisory committee meeting set for April at the Bureau's headquarters in Suitland, Md., according to assistant director of the Decennial Census Dan Weinburg. "For 2020 we're thinking about a sea change in the way we collect data," he says. "Right now, you have to fill out a form and send it in, and if you don't we'll send someone to your house. We want to make as many ways to respond as possible." Weinburg says the Bureau hopes to maximize self-response by developing "a generic methodology that would work on almost any device," including smart phones, which are predicted to be ubiquitous in the U.S. a decade from now. To keep costs down, "we have to think dramatically about transitions technology," he says.

Planning for the 2020 count started in 2008. Weinburg says that going completely electronic by then is unlikely because of the amount of controlled experimentation that has to be done before the change could be made. But he hopes to keep the cost per household on par with 2010 until a new, cheaper strategy can be implemented.

Smaller samples

Another obvious way to cut costs would be to count a smaller sample and extrapolate the data to the greater population—a common practice in the social sciences. This method forms the basis of the Census Bureau's American Community Survey—a more detailed form sent annually to a representative sample that replaced the long-form decennial census survey that used to go out to one in six households.

But sampling introduces a new type of error on top of the existing ones (omissions, nonresponses and measurement errors), says David Swanson, chair of the Census Advisory Committee of Professional Associations, which advises the Census Bureau on issues related to improving the accuracy and reliability of the enumeration process. Besides, in mandating the census, the Constitution states that congressional districts have to be apportioned according to population, so actual enumeration (even if it's not door to door) has to occur.

The data are already collected

Demographers also are toying with whether the Bureau should take advantage of existing records, such as social security, tax returns and draft records. This would eliminate the nonresponse problem, but would have to be supplemented with enumeration for those who are not on record with the government.

At the 2010 American Statistical Association's Joint Statistical Meeting to be held this summer in Vancouver, demographers will pitch new ideas for data collection in a session called "What if 2010 was the last census?" Swanson and New York City's chief demographer Joseph Salvo have been invited to share their thoughts on the topic. They both propose using some form of existing documentation. "We have to look for different lists that are already out there," Salvo says. "If a person is part of that list, that might be their way of responding."

Using government records for demographic information is a controversial idea, not only because enumeration is mandated by law but also because people are skeptical about trusting the government to protect their privacy. "We have to talk to people about whether we could use that information," Weinburg says.

Could taking advantage of existing records retire the census, as we know it, for good? Swanson says that combining the Master Address File (a regularly updated map of every household) with preexisting documents would yield about 85 percent of the required data. The remaining 15 percent would come from what Swanson calls his "secret sauce," which he'll reveal during his presentation this summer, and no sooner. "This is how a radically different census could be done," he says.