We try not to get too data wonky on Minnesota Compass. Our typical users are people who are looking for reliable data and information in one place, maybe to learn about their community, for a grant application, for community advocacy, or for a school assignment. You’re not usually interested in how the data came to be or how it got to you. Rather, you trust our Minnesota Compass researchers to assemble accurate and reliable data (Thank you for that trust!).
In this article, I’m going to talk about some data-wonky things, but hopefully in a friendly and approachable way. And it matters to you because it’s all about trust: the trust you place in us – and that we place in our data sources – that information is accurate and reliable. So we’re paying attention to an issue called “differential privacy,” which could affect the degree to which we can trust data and estimates coming out of the U.S. Census Bureau.
What is differential privacy?
When you participate in Census Bureau programs by completing forms and surveys – including the decennial census – you are assured that your information is protected and confidential. That’s important. Would you be truthful and honest on those forms if your information wasn’t protected? Would you complete those surveys at all? Probably not.
In the interest of confidentiality, the Census Bureau announced a new policy of releasing only “differentially private” data from the 2020 Census. This means that they intentionally add errors into most data below the state level (like cities and towns) and when data are split out by characteristics like race, income, or age. The idea is that this would further protect respondents’ confidentiality by making it even more difficult to identify individuals in Census Bureau datasets.
Sounds okay, right?
What are potential issues with differential privacy?
Census Bureau products – like data from the decennial census and American Community Surveys – are used extensively by government officials, planners, scientists, and others for decision-making, funding, planning, policy-making, and research. Intentionally introducing error means that we may question whether estimates are still trustworthy.
And we should question whether differentially private data are trustworthy.
Researchers on the IPUMS project, with collaborators from across the country, compared 2010 differentially private demonstration data and published 2010 data (without differential privacy). Their results show that differential privacy has negative impacts on the reliability of estimates, especially for small geographies, tribal lands, children, Black communities, and Hispanic/Latino communities.
In addition to issues with reliability, others have noted that differential privacy will continue to result in:
- Significant delays in releasing data due to the time it takes to introduce error
- Potential lack of equity in introducing error into data for different areas (e.g., rural/urban)
- The absence of detailed guidance, inappropriate use of differentially private data
Why should I care?
Chances are, if you’re a Minnesota Compass user, you already know how valuable census data are to telling our stories. But I’ll start there anyway. Whether you notice it or not, data from Census Bureau products are used to make decisions and influence policy about your life. To name a few, these data are used to predict school enrollment in specific communities; to make decisions about where to place a new business, transit line, or assisted living facility; and to understand who is affected by a natural disaster. Reliable data are critical to having a sound understanding of our communities and making well-informed decisions.
Differential privacy potentially threatens what we reliably know about ourselves and our communities. This is particularly the case for smaller communities and sparsely populated areas, American Indian communities and tribal lands, Black and Hispanic/Latino residents, and children.
If I’m concerned, what can I do?
- You can stay informed. IPUMS has been heavily involved in the issue of differential privacy, from keeping users informed to conducting research on its impacts. You can visit their website and sign up for their mailing list to receive updates on anticipated changes to Census Bureau products.
- You can use your voice. U.S. Census Bureau Director Robert Santos is interested in feedback on how to balance protecting confidentiality and ensuring census data products are usable and reliable. The Federal State Cooperative on Population Estimates Steering Committee is advocating against differential privacy in future data products, including Population Estimates (which we use almost exclusively for our demographic estimates on Compass) and the American Community Survey (which we use extensively in topic areas across Compass). If you agree with their points, you can sign on to a letter to Director Santos that will be sent during the week of August 15, 2022.