Using Administrative Data to Combat Statistical Erasure

While large-scale survey data have yielded important understandings about the economic and social conditions that structure individual and family well-being for large population groups in the US, these data often share one weakness: probability sampling does not yield sufficient sample sizes of smaller populations. This is a particular issue for American Indians/Alaska Natives (AI/AN) and Asian-Americans, all of whom get routinely collapsed in the category of "other" in national data, a practice that scholars from these communities have long critiqued as "statistical genocide" or "involuntary invisibility."

While data that contain sufficient counts of AI/AN and Asian Americans are lamentably uncommon, more fine-grained data are even rarer. Combining all AI/AN persons into one racial category means flattening over 500 diverse nations with unique histories and cultures into one seemingly homogenous category. Similarly, lumping Asian Americans into one category can obscure large within-group differences, particularly when Pacific Islanders are also included. American legacies of settler colonialism and imperialism further appear in the Census combination of Native Hawaiians and Pacific Islanders, which obscures heterogeneous experiences of pre-colonial and post-colonial histories.

This roundtable will present new thinking about how to use integrated administrative data to capture the experiences of smaller populations, with specific consideration of American Indian and Pacific Islander representation.

Administrative data arises from persons' interactions with government agencies, creating records that can be transformed into data that contain large observation counts (populations rather than samples) over long periods of time. Integrating multiple administrative data sources by merging at the individual or household level can yield accurate reports on earnings, transfer income, voting, residential addresses, and other factors relevant to studies of human populations. For the cost of a modest survey of a few hundred households, data scientists can assemble a full-population data source with millions of observations.

The organizer will provide a framework for thinking about administrative data use for studies of small populations, including ethical and legal considerations (indigenous data sovereignty rights and the CARE Principles for use of tribal data. Then, three emerging scholars will share their experiences with developing and using administrative data. First, The first presenter will speak about her research using administrative records to understand economic well-being around a birth for AN/AN birthing parents. This work, done in collaboration with community partners, will lead to new understanding about how the Washington State Paid Family and Medical Leave policy affects AI/AN families. The second presenter will discuss possible methodological innovations for identifying Pasifika peoples within state administrative data, as well as ways in which to involve affected small-population communities in data use agreements. Finally, the third presenter will speak about her work with the Rhode Island Data Ecosystem using integrated datasets to analyze metrics of quality mental health service for children and youth and the tensions between conventual use of major ethnoracial groups and the use of more disaggregated approaches. Cross-cutting conversations will focus on practical and ethical considerations of working with full-population data for small populations.