Can Administrative Records Be Used to Reduce Nonresponse Bias?

Abstract

One option for addressing the bias that may result from survey nonresponse is to make greater use of the administrative records that federal and state agencies compile. Such records have been used to assess response bias but less often to correct for such error. Direct substitution of administrative records for survey data, as is done for income data in Canada, provides a means of compensating for survey nonresponse; but the limitations of such data must be recognized. Administrative records may not cover the entire population of interest, may utilize a different unit of observation, may have wide variation in data quality across items or by agency, and may have timeliness issues. In using administrative records, researchers cede control over the content of individual variables, which may differ from survey concepts and be subject to change. Furthermore, privacy protections embodied in law restrict the use of many types of administrative records.