Letter from the Director | Fall 2005
Labor Day
Labor Day: or the best and the worst of times. On the one hand, students are returning to campus, bringing excitement and possibility. On the other hand, our seemingly realistic summer research plans are revealed to have been too ambitious; projects are shelved as the rush of preparing courses commences. The worst is that the start of the endless meetings is lurking around the corner. ISERP looks different. We have a new logo. The 8th floor computer lab, built just on the edge of the laptop revolution, is gone. Our core staff, scattered here and there on the 8th floor, are gathered (perhaps uncomfortably) together. QMSS has moved from the 3rd floor to a more central location on the 8th, and so on. In good news, the reason for most of these changes is continued growth.
Over the past two decades, fueled by changes in social science research strategies and increased computational power, social scientists have begun to recognize that threats to confidentiality through deductive disclosure of respondent identity are real. First, effective research design rests more and more on clustered samples enabling exploitation of advanced multi-level, network, and spatial analysis techniques that allow us to meaningfully locate respondents in the contexts that shape their opportunities, values, and experiences. At the same time, enhanced computation capacity means that it is a piece of cake to cross-classify attributes of individuals in order to reveal identities. For respondents, this means if someone knows you are in a sample, has access to the dataset and a computer, they can find you-and all your responses to the questions you were asked. Individuals in clustered samples, schools, organizations, neighborhoods, and so on are at greater risk simply because more people know that a study in one of those settings took place. Respondents might well worry; social scientists need to worry. One threat is messy but less serious: the threat arising from individuals seeking to know something about someone in a study for purely private reasons-spouse, daughter, co-worker, and so on. A second threat is more serious. Those hostile to social science research, especially those hostile to government-sponsored social science research, have only to reveal the identities of individuals and their confidential responses to sensitive questions to "take down" the enterprise. Nor should such people need to care for subtlety. Individuals may care about false positives, but false positives are not really less problematic than exact matches for those seeking to delegitimize the whole apparatus of social science research.
Protection against deductive disclosure of respondent identity does not come easily. One path lies in restricting access to data to legitimate researchers whose activities are governed by academic norms. Here, Institutional Review Boards (IRBs) can play a major role, and many studies now require IRB approval for security and data-handling protocols before release, for example the PSID and Adolescent Health. Restricting access to data to those with IRB approval means that ordinary people may lack opportunity to study on their own. As with all restrictions of this kind, "public release" means something else. But this seems a small price to pay for protecting respondents. Dads seeking to learn about daughters may be stymied by such restrictions, but those with animus to social science can easily get around them. To stem their efforts significantly, more stringent steps are required. The most obvious, and at the same time the most bothersome, is dusting data.
All of the steps one can take may be to no avail if individuals lose faith in confidentiality. In this newsletter issue, Ken Prewitt, who directed the census in 2000 and now serves as chair of the Department of International Affairs, considers privacy, statistics, and administrative data. We are sure that his comments will stimulate much thought.
We hope you check out the new ISERP configuration. As usual, send along comments and ideas for future issues.
Peter Bearman, Director
Institute of Social and Economic
Research and Policy
