Report Publish Date: August 2021
This report describes a landscape scan of the ethics involved in imputation, a practice by which researchers add racial and ethnic identifiers that are created through statistical calculations, onto a dataset where it had been omitted so they can disaggregate data by race.
The authors found that:
Disaggregating data by race and ethnicity helps researchers understand and describe the racialized systems at work in our society. But race and ethnicity are not always included in data sets.
Imputation, or adding racial and ethnic identifiers to datasets using statistical calculations, can aid researchers who want to disaggregate data. The authors of this report wanted to investigate the ethical considerations that come into play when researchers use imputation.
The authors reviewed the literature on imputation and interviewed experts on data disaggregation, imputation, and racial equity.
"[I]mputation ... intersects with big-data systems, as big data are often used as the foundation for, or to train inputs into, predictive statistical models.”
—K. Steven Brown, Yipeng Su, Jahnavi Jagganath, Jacqueline Rayfield, and Megan Randall, Urban Institute
“[R]esearchers should collaborate with community members and other accountability partners to decide whether imputation is the right approach to filling missing data and whether the risks outweigh the benefits,” the authors write.
A growing body of literature has demonstrated how seemingly “race-neutral” policies and systems can reinforce white privilege. Alongside historical research about how racist policymaking has affected people of color in US communities, disaggregation of data by race and ethnicity is a critical tool for shining a light on the racialized systems at work in our society. There are strong ethical and practical reasons for disaggregating data: disaggregated data can provide clarity in areas where disparities have been suspected but not identified. Moreover, disaggregating data enables people to see themselves reflected in data, which in turns enables them to make decisions, inform policy, and advocate for a more just and equitable distribution of resources.
Despite the benefits of disaggregated data, many high-value datasets lack information on race and ethnicity. The absence of disaggregated data has harmed communities of color and obscured disparities in health, wealth and financial well-being, justice involvement, safety net benefits, and other policy domains.
In response, data scientists and researchers have developed, and continue to expand, creative methods for appending race and ethnicity onto datasets lacking those data, allowing policymakers to disaggregate those data and track racial disparities to inform policymaking. These methods are often held up as alternatives to the “gold standard” of collecting original, self-reported data on people’s race and ethnicity when those original data are not feasible to collect (such as with historical data) or are not allowed to be collected (such as with tax data).
In this report, we focus on risks associated with a method known as imputation, primarily because it is increasingly employed to append race and ethnicity onto new datasets. Although imputation is being used more often, the literature on associated ethical concerns remains sparse.
Urban Institute, August 2021
This study and report was conducted and created by the following people.
Gathering and sharing accurate data can help communities, policymakers, journalists, and others understand the barriers to health some face and take action to overcome them. Explore public health data for analysis of the many factors that shape health.
