Econometric Modelling Using Factual Anonymized Data
Over the last decades empirical research in the social sciences showed an
increasing interest in the analysis of microdata. Due to confidentiality
and privacy considerations, the original data can often not be made
available to the researcher. Individual and household data as well as
firm level data contain sensitive information on the observational unit,
whose confidentiality has to be protected against disclosure in the
interest of the observational unit, but also in the interest of the
data collecting institution. Private and public data collecting
institutions therefore become interested in the provision of scientific-use-files
that optimally combine the interests of the survey respondents, the data
collecting institution and the academic user. In order to minimize the
probability of disclosing individual information, statistical offices and
other data collecting institutions apply various masking procedures.
However, common data mashing procedures severely reduce the quality of
the data by reducing the efficiency of econometric estimators and, even
worse, by rendering estimators to be biased and inconsistent.
The goal of this research project is to analyze the consequences of popular
disclosure limitation methods on the properties of nonlinear estimators.
This research project investigates to what extent appropriate econometric
techniques can obtain parameter estimates of the true data generating process,
if the data are masked by noise addition or blanking.