Disparate impact and reading data responsibly

Disparate impact is a legal concept that is useful for leaders to understand at a working level. Title VI of the Civil Rights Act prohibits discrimination based on race, color or national origin in programs that receive federal funds, and federal regulations also reach criteria or methods of administration that have discriminatory effects. The U.S. Department of Justice Title VI Legal Manual describes the analysis in three parts. First, is there an adverse disparate impact: a specific policy that causes a harm falling disproportionately on a protected group? Second, does the recipient have a substantial legitimate justification: is the policy necessary to meet an important, integral goal, supported by evidence rather than speculation? Third, is there a less discriminatory alternative that would meet the same goal?

You do not need to be a lawyer to use these three questions as a design discipline. They are the same questions a careful program team would ask anyway: What is the effect? Why do we need this? Is there another way? What you should not do is announce a legal conclusion. Whether a practice violates Title VI, the Minnesota Human Rights Act or any other law is a formal determination for legal counsel, the Equal Opportunity and Access Division, or an enforcement agency. Your job in an equity review is to surface the pattern, document the reasoning and route concerns to the right office.

Data is where many reviews go wrong. Public sources such as MN Compass, Minnesota Department of Health equity reports, and DHS program dashboards can show disparities by race and ethnicity. They cannot, on their own, tell you why a disparity exists. Before using a number, check its date, its definitions, which groups were combined, how small the counts are, and whether data is missing more often for some groups. A combined category can hide large differences between, for example, Somali, Hmong, Karen and Ethiopian Minnesotans. Small numbers can swing sharply from year to year. Never invent or estimate a figure to make a point, and always cite where a number came from.

Numbers describe patterns; people describe how the pattern is lived. The GARE toolkit pairs data with community engagement for this reason. Talk with the people and partners affected, including counties and Tribal Nations where they administer services, about what the process is like from their side. Their account often points to the mechanism that the data only hints at.

Fictional case: Grace, a data analyst, finds that denials for a service authorization are more common in two racial groups. Her manager, Olu, is drafting a briefing for leadership. A colleague suggests titling the briefing 'Evidence of discrimination in authorizations.' Grace notes that the counts for one group are small, that race is missing for many records, and that the team has not yet looked at which denial reasons drive the pattern.

Find one published Minnesota disparity figure relevant to your area, from MN Compass, MDH or a DHS report. Write down its source, date, definitions, which groups are combined and one thing it cannot tell you. Share the note with a colleague who uses the same data.

When you have seen a disparity number used at work, how often was its date, definition or data gap mentioned? What would help you ask about those limits comfortably?