Data From the Retirement History Survey

GAO initiated a study on retirement patterns among the general workforce and attempted to use the Retirement History Survey (RHS). RHS consists of data collected by the Bureau of the Census on approximately 12,000 individuals who were followed by the Bureau into and through the retirement years.

The study was terminated because of the poor quality of the RHS data. Overall, GAO encountered nine kinds of problems: (1) inconsistencies in items across years that prevent consistent analysis across years; (2) inconsistencies within individual questionnaires; (3) inconsistent response categories across years; (4) unrealistic extreme values; for example, there are respondents who indicated that they have 44 or 99 children; (5) frequent instances of missing values and inconsistencies across years in the number of missing cases for the same item; (6) improperly defined response categories; (7) logically impossible negative values for items such as income; (8) internal errors across variables within years; and (9) inconsistencies in responses across years. These kinds of problems differed in the frequency with which they occurred and their possible damage to any particular analysis. GAO believes that potential users of RHS data may be better served if they know the condition of the data before committing significant funds and effort to acquire the data tapes.