The Dark Side of Numbers: The Role of Population Data Systems in Human Rights Abuses

Social Research, Summer, 2001 by William Seltzer, Margo Anderson

Methodological and Technological Safeguards

Even if items and classifications that define one or more target populations are included in a national data system, a range of methodological and technological procedures can be used to reduce the potential negative impact of such inclusion. For example, if the data system is based on sample rather than full-count data gathering, the resulting information is of little help in providing microdata that can be used to provide operational lists of the members of a target population. Depending on the size and type of sample, the results may also be of limited usefulness in providing operationally relevant mesodata. Typically, even relatively large-scale national sample surveys based on multistage samples of clustered households would be of limited usefulness in this regard. On the other hand, essentially unclustered systematic samples of census enumeration records or population registers might provide operationally useful mesodata if results were shown for small geographic areas.

Another broad technological approach is the deliberate introduction of errors into the data set. These include systematically swapping responses for individual items between records, introducing perturbations in specific items, top (or bottom) coding of quantitative items so that unduly large (or small) responses are grouped together to protect the identity of respondents, coding categorical data in broad response categories, or using only large areal units for similar purposes. As discussed earlier with respect to the elimination of variables, reducing the level of substantive or geographic detail available also reduces the usefulness of the data for certain users and uses.(1)

Organizational and Operational Safeguards

Suitable organizational and operational arrangements have also been used to help protect against the misuse of population data systems, although to date these arrangements have not been systematically discussed. For example, in the Netherlands the population registration system is deliberately kept decentralized (Begeer 1998); in the United States decisions at the Census Bureau related to the release of data that may pose confidentiality issues are made by a committee that is independent of both the concerned substantive and processing divisions; and in several countries the machine-readable census-data files are stripped of most or all individual personal or exact address identifiers. More complex procedures have been used in some sample surveys that collect sensitive data. In one case three files were established: an anonymous data file, an identifier file, and a "bridge" file that provided the link between the other two files, with the bridge file kept in a foreign country immune from domestic court orders. As with other safeguards, the degree of protection afforded by such operational and organizational arrangements is rarely absolute, particularly with respect to threats posed by the misuse of mesodata. Nevertheless, the use of such safeguards--jointly with other approaches--can make misuse more difficult and thus deserve more careful attention.