Limitation Methods for Protecting the Con ® dentiality of Statistical Data

The protection of con®dentiality is both a private and a public issue. Corporations collect extensive information on customers, some of which is highly sensitive and often protected by law. Similarly government agencies collect both administrative and statistical data subject to pledges of con®dentiality and as more and more administrative data are used for statistical purposes there are increased dangers of disclosure of con®dential data. At the same time, the public demand for data from statistical of®ces about diverse aspects of modern society seems insatiable. Other government agencies use statistical data for the allocation of funds and the monitoring of social programs, policy analysts use statistical data to do calculations. The potential effects of new legislation have to be investigated, and academic researchers are constantly looking for data to validate and extend theoretical social science models. Even Saskia Groenwald, Otto Normalverbraucher, Jan Modaal, John Smith ± or whatever their names in countries around the world: they are all showered with statistical data on crime, money, employment, and health, via the news-media to which they are exposed. And their lives are to a large extent governed by policies that are fueled by the analysis of data collected by statistical agencies of the countries in which they live. In a way, statistical agencies provide mirrors for society, and society itself is a keen user of its own images, just to gaze at in satisfaction, amazement, embarrassment or in order to improve its looks. The images are the data that a statistical of®ce releases, after having collected and processed them. Of course, the metaphor of mirror image is simply that, a metaphor, and while it is evocative, statistical data do not quite provide an accurate picture of reality. First, forming an image of an aspect of society is not as easy as re ̄ecting light in a mirror. Second, the statistical image lacks detail. This lack of detail is in part a ®scal necessity, since it is more costly to employ sophisticated methodologies (or models), which produce better images