Problems with Critical Government Data: Evidence from the Death Registry

• In Event: Tackling Biases in Tricky Data Collection

Abstract

To function effectively, governments need high quality data. Collecting these data, however, often requires delegation of responsibility to multiple agents, risking biased information due to differences in agents’ capacity, competence, or preferences. As a result, even “high quality” data can have systematic shortcomings, as we illustrate in analyses of data from the U.S. death registry, an administrative data set built upon death certificates issued by county officials. Our study shows that coarsening, heaping, and errors arise systematically due to limited county government resources---data from smaller, poorer counties are of a significantly lower quality than data from larger, richer counties. These heretofore undocumented problems with the registry engender inequities in our understanding of Americans’ health outcomes and our ability to improve upon them. Given the imperfections of administrative data, credible inferences can only be obtained if we devote careful attention to the political economy of data to better understand and account for these limitations.

Authors

• Scott Cook, Texas A&M University

• David Fortunato, University of California, San Diego