Comparative Study of the Effectiveness of Perturbative Methods for Creating Official Microdata in Japan

摘要

The Statistics Bureau of Japan currently creates and provides anonymized microdata for six surveys. Anonymized microdata from the Population Census of Japan are created using non-perturbative methods such as recoding, top coding, and record deletion as well as the perturbative method of swapping. This paper analyzes several types of anonymized microdata created based on individual data from the Population Census, and explores the potential of using perturbative methods such as swapping and PRAM to create anonymized microdata from Japanese Census Data. Results suggest that perturbative methods can increase data quality, but should be selected according to the properties of the microdata that are to be anonymized. Perturbative methods have the potential to help further enhance statistical methodologies in Japan.