The future of the U.S. Census — the data set many Americans use to judge themselves and their compatriots — is in doubt, according to participants on last night’s panel “Censuses and Surveys: Still Useful for the Common Good?”

Enlarge
#gallery http://daily.stanford.edu/image/full/7396
Jason Chuang

three speakers--Kenneth Prewitt, Douglas Rivers, Henry Brady(from left to right) --discuss the practicality and usefulness of censuses and surveys in today's social science research and public policy efforts in a presentation called "Censuses and Surveys: Still useful for the common good?"

Concerns about the privacy of personal information plagued the 2000 U.S. Census, according to panelists, and this public distrust led to a precipitous drop in the Census response rate, which called the accuracy of government statistics into question.

In order to remedy this erosion of data quality, social scientists may soon be forced to dig into other government agencies’ databases, such as the Internal Revenue Service or Social Security Administration, last night’s speakers said. This data, which contains sensitive information, must be handled carefully in order to avoid raising more privacy concerns.

The panelists at last night’s event in the Green Library Social Sciences Reading Room included Columbia Prof. Kenneth Prewitt, former director of the United States Census Bureau; UC-Berkeley Prof. Henry Brady, head of that school’s Survey Research Center; and Political Science Prof. Douglas Rivers, founder and CEO of the Internet-based survey company Polimetrix.

Prewitt estimated that about 70 to 80 percent of what he calls the “national statistical system” is based on survey data, but said he thinks this number will be halved in 20 years.

“Why is that?” he said. “The answer is almost always the increasing cost of survey data. The sample cost of the 2010 Census will be close to $12 billion, up by five billion dollars in the past decade. That is driven by how hard it is to get people to answer questions. There is across the survey industry a response rate crisis.”

Prewitt added that incomplete questionnaires are increasing at “astronomical rates.”

Brady said a convergence of factors was responsible for the decline in response rates.

“Marketers haven’t helped us at all with their incessant pestering of people. I think trust in institutions has gone down,” he said. “Technology hasn’t helped us that much with people that now have caller ID and if they don’t know the number that’s calling, they don’t return the call. Or they have answering machines and if they don’t like who you are, they don’t reply.”

Rivers hypothesized that these factors might be related to a general decline in social cooperation.

“Something like the Census that has typically had extremely high levels of voluntary cooperation is going down,” he said. “That’s difficult to explain, other than higher levels of social distrust.”

Much of the discussion revolved around confidentiality issues inherent in using administrative data for statistical purposes — issues that Prewitt outlined in detail.

“I spent hours and hours trying to explain to people ‘send in your Census form, because after you do, we don’t care about you. You just become a statistical record,’” he said. “But administrative data are primarily about particular individuals. When the federal government uses administrative data to manage Social Security they actually want to know who people are, what their income history and ages are and so forth.”

In the end, Prewitt said, the blurring of the line between administrative data — like Social Security information — and statistical data would likely lead to a decrease in privacy and a further decrease in public participation in surveys.

“We’re not surprised when administrative data keeps our identity attached to it,” he said. “But for statistical data, we don’t care about people’s identities. Therefore, we have drawn a firewall between statistical data and administrative data. Now that firewall is being breached. It has to be. We’re going to have a period with much less privacy and confidentiality. I cannot tell whether we will get a public outcry when these datasets are merged. I just don’t know.”