Effective record linkage for mining campaign contribution data

摘要

Up to now, most campaign contribution data have been reported at the level of the donation. While these are interesting, one often needs to have information at the level of the donor. Obtaining information at that level is difficult as there is neither a unique repository of donations nor any standard across existing repositories. In order to more meaningfully mine campaign contribution data, political scientists need an accurate way of grouping, or linking, together donations made by the same donor. In this paper, we describe a record linkage technique that is applicable to various sources and across large geographical areas. We show how it may be effectively applied in the context of nationwide donation data and report on new, previously unattainable results about campaign contributors in the 2007–2008 US election cycle.