Disclosure Risk from Factor Scores in a Remote Access Environment

Remote access is a promising tool for broadening the access to microdata without violating confidentiality requirements. In a remote access setting the user submits queries to a system provided by the statistical agency and only the results of the queries are reported back to the user. Since no direct access to the data is granted, generally no alteration of the underlying microdata is required. Still, remote access bears the risk of disclosing sensitive information even though the actual data are not directly available. Most disclosive queries are easily detected and can be suppressed by the system. However, more complex procedures such as multivariate analyses can also lead to a breach of confidentiality if applied in a sophisticated manner to exploit certain features of the data. In this paper we illustrate how an intruder could employ commonly used factor analysis to obtain sensitive information regarding the underlying microdata. We present the general concept and evaluate the approach using a German establishment survey, the IAB Establishment Panel. We find theoretical and empirical evidence for a high risk of disclosure from factor analysis.