This paper addresses the issue of long-term preservation of and access to digital government information. We show how the preservation process is enhanced by storing an infrastructure-independent representation of the raw data, together with a model dependency graph (an executable graph of database view mappings). This allows for the design of decision-support tools and services for improving government transparency and promoting citizen access to eGOV data. A case-study, the Florida Ballots Project, is used to illustrate the approach.