Comment: The Future of Administrative Records in the Census Bureau's Demographic Activities

Abstract

I will focus my remarks on demographic activities at the U.S. Bureau of the Census. This may be somewhat provincial, but since the Census Bureau is often considered a center of action-or inaction-in using administrative records, I will discuss goals 1 and 4 particularly. Goal 1 calls for expanded use of administrative record systems in the conduct and evaluation of decennial population censuses and in current population estimates. Goal 4 promotes expanded use of administrative records in all phases of household surveys. I will begin with a brief review of how the Census Bureau's demographic activities have benefited from use of administrative records. Then I will outline some difficulties that attend their continued or expanded use. I will then describe some future uses of administrative records that we are now planning or considering. All of this suggests some guidelines for future use. Many important demographic activities at the Census Bureau have benefited substantially from our statistical uses of administrative records. Here are a few recent examples. In 1980 we used a sample of persons on individual tax records matched to census records to evaluate coverage of the decennial census. We also used administrative records to identify special target populations in the census for improving its coverage. For the Income Survey Development Program (ISDP), we matched with a variety of administrative records. We also used federal and state administrative record systems as sampling frames for pilot surveys associated with the ISDP. The Current Population Survey (CPS) and other current demographic surveys depend fundamentally on administrative records: local building permits are the basis for updating the CPS address sample for new construction between censuses. Finally, there is the Census Bureau's intercensal estimates program, which produces estimates of the population of all states, counties, and units of local governments. Federal, state, and local administrative records of various sorts document the births, deaths, and migrations that drive the changes in these intercensal estimates from year to year. To understand the Census Bureau's plans for using administrative records in the demographic area in the future, it is useful to review briefly the problems in such use. There are technical, organizational, and perceptual problems. Among the technical problems we confronted in using such records in 1980 Census processes w re lack of residential addresses, geocoding problems in rural areas, lack of accurate apartment identification in multiunit structures, and matching difficulties in the absence of Social Security numbers in census records. More generally, most problems in our use of administrative records arise because such records are collected and maintained for particular administrative purposes that rarely match our statistical purposes. Concepts and measurement periods are rarely the same. Quality control is sometimes insufficient. In addition, the coverage of administrative record systems is usually not coincident with the coverage of the statistical systems with which they might be used. Other problems arise because the program agencies that collect such records sometimes change them, or fail to change them, without regard to our needs. One difficulty is communication. Sometimes the program agency does not inform us of such changes. Even when communications are good, the problems introduced by changes can be intractable. That is, it is sometimes impossible to distinguish data changes in the demographic or economic characteristics measured by the administrative record system. These problems are no one's fault. They result when agencies with very different purposes collect data to meet their particular needs, then seek efficiencies by combining the data for statistical purposes. The more integrated the uses of administrative records in statistical systems, the more troublesome these difficulties will become. Nonetheless, administrative records play a key role in the identification of errors, in imputation pro-