ABSTRACT
The storage and retrieval of microdata involve serious problems because of the complex documentation required. If storage is costly, and keeping archives in the best storage involves repeated costs for updating, we need some criteria for deciding what to keep most easily accessible, what to keep in various states of accessibility, and what to store or even to destroy. Data from systematic official records or probability samples are more likely to provide clean comparisons than are non-sample data, or data from small sub-populations. Good interviewing or data collection methods are more likely to be reproducible than are bad ones. Storage of data has a number of costs some intermittent, and some associated with each use. There are investment costs in preparing data and documentation for archives, and some strategy decisions. If a major purpose of archiving microdata is to facilitate study of the natural experiments in history then we need the data available in their most flexible form.
