Merging time-series Australian data across databases: Challenges and solutions

Abstract

This study discusses the differences in company identification across sources of Australian data and raises important issues which should be considered prior to merging across databases. In particular, we show that the practice among accounting databases of overwriting prior identifiers used by a given company, with its most recent, results in failure to match data which actually exists. We suggest a method for reconciling these differences and show that our method results in a match rate of 97 percent with the Aspect company identification file, and 94 percent after missing accounting data is considered. This contrasts with a match rate of only 71 percent when performing a direct merge.