Ex-Census chief proposes new privacy protection

The former director of the Census Bureau has proposed that future population surveys no longer release data for geographical units such as street blocks. The proposal--in a June 26 letter to the White House's top statistician--comes at a time of increasing concern about the privacy of government-collected data and the commercial marketing databases that use such data. Last month, for instance, Rep. Tom Sawyer, D-Ohio, introduced a bill, H.R. 2136, that would extend the privacy protections currently in place for the Census Bureau data to more than 70 federal agencies that collect individual and personal data. In a letter last month to Katherine Wallman, the chief statistician at the Office of Management and Budget, Kenneth Prewitt, the Clinton administration's Census Bureau director, said because of the high error rate of block-level data, releasing it has heightened the politically charged issue of whether the census count should be statistically adjusted through a process known as sampling. "The debate [over sampling] does not disappear but might proceed more scientifically if the smallest geographic unit for reporting [the] 2010 decennial census was, for example, the census tract," said Prewitt, who frequently clashed with House Government Reform Census Subcommittee Chairman Dan Miller, R-Fla., and other Republicans on the panel. Once all 2000 census data are released later this year, individuals can obtain statistical profiles of the number of people in their block, the number of a given race and the number of people earning certain income levels through a portion of the bureau's Web site. "Reporting data for small geographic areas, such as census blocks, cannot but exacerbate public anxieties about confidentiality," said Prewitt. He attributed much of the privacy concerns to what he described as "a misuse of data by the commercial sector" rather than to the bureau, which he said operated under strict confidentiality rules. Prewitt said a procedure known as "data swapping," a method of modifying the publicly available block-level data to protect the confidentiality of individuals who otherwise would match a profile in such a small geographic area, provides additional privacy protection. But such "disclosure avoidance" methods are hard to explain to a general audience. "The disclosure avoidance procedures to which block-level data are presently and properly subject are not well understood outside the statistical community," Prewitt said. "My experience is that trying to explain them heightens more than allays suspicions." Elaborating on his letter in remarks to a gathering of the Washington Statistical Society last month, Prewitt said privacy and confidentiality was one of the top four issues raised by the 2000 census, the other three being the inclusion of a mixed-race option, whether to count Americans abroad and the continuing controversy over whether to adjust the census counts. "The confidentiality issue is just very big and is going to get bigger," he said.