On Regression-Tree-Based Synthetic Data Methods for Business Data

The challenge of balancing the competing objectives of allowing statistical analysis of confidential data and maintaining confidentiality is of great interest to national statistical agencies and other data custodians seeking to make their data available for research. This balance is often characterised as a trade-off between disclosure risk and data utility, where disclosure risk attempts to capture the probability of a data release resulting in a disclosure, while data utility attempts to capture some measure of the usefulness of the released data, see [6]. To date, most of the literature on addressing this balance has focussed on data about individuals, however, the same problem arises in the context of data about businesses and enterprises. It is the purpose of this paper to provide an empirical evaluation of existing methodology for individual data being applied to business data.