Summary
Aim
This bulletin describes the results of a validation technique the Bureau applied to test the accuracy of the matching processes that underpin the reoffending database.
Abstract
In 2001, the NSW Bureau of Crime Statistics and Research developed a reoffending database. This database links individuals' criminal court appearance records over time and so enables the measurement of recidivism in NSW. This bulletin describes the results of a validation technique we applied to test the accuracy of the matching processes that underpin the reoffending database. We first describe the development of the reoffending database and the deterministic matching criteria that are its foundation. We then describe the validation technique of applying those deterministic matching criteria to a dataset of individuals with known identities and measuring the number of false positives and false negatives that resulted. The validation results suggest that the error rates in the reoffending database are likely to be acceptably low. Finally we discuss the testing of an additional matching criterion involving residential postcode and recommend its adoption.