Commentary: Future U.S. National Statistics Use of Administrative Data

Abstract

Government statisticians are intrigued by the possibility of accessing administrative data as a way to enhance survey and census data. Survey data are rich in attributes, but they exhibit sampling errors. More consequentially, they demonstrate non-sampling errors, often because of nonresponse, but also because of responses that are incomplete or inaccurate, frequently due to the respondent’s inability to recall. Additionally, panel data are subject to attrition. Administrative data can help compensate for these problems importantly because administrative data, such as social security files, include almost everybody. Further, supplementing statistical data can be cost-effective, because an enormous amount of administrative data are already collected to support the functional operations of government agencies, and so are potentially available. In many instances, respondent burden in surveys can be lessened because certain attribute values may be available from administrative records. Some U.S. federal agencies, e.g., the Social Security Administration (McNabb et al., 2009) have linked administrative data with survey data to broaden its demographic and socioeconomic measures and also to improve the quality of the survey data. See National Research Council (2005), pp.45–48 for additional examples and discussion of benefits of such data linkage.