The Coalition on Human Needs, Partnership for America’s Children, and Zero to Three had planned to submit these comments to the spring 2025 Census Scientific Advisory Committee meeting urging the Bureau to make sure when it adopts a new privacy protocol for the American Community Survey (ACS) that it preserves the ACS ability to produce accurate state and local child poverty data. Child poverty is the most important single indicator of child well-being, but the privacy approach used in the 2020 census would make it impossible to produce child poverty data from the ACS. Unfortunately, the meeting, which was scheduled to be held March 13 and 14, was cancelled and the Committee itself was terminated. We are therefore posting our planned comments here instead.
We submit these comments on behalf of the Coalition on Human Needs, Partnership for America’s Children, and Zero to Three.
The Coalition on Human Needs is an alliance of national organizations working together to promote public policies which address the needs of low-income and other vulnerable populations. The Coalition’s members include civil rights, religious, labor, and professional organizations, service providers and those concerned with the wellbeing of children, women, the elderly, and people with disabilities. The Coalition on Human Needs monitors and tracks data on human needs in the United States, including data on poverty, on policies that reduce poverty, and on hardship. The Coalition uses Census data including the American Community Survey (ACS) in its work. Along with the Partnership for America’s Children, First Focus on Children, and Zero to Three, the Coalition is one of four organizations that formed and continues to co-lead Count All Kids, a national group of child-serving organizations that is working to improve the count of young children in all Census Bureau demographic products.
The mission of the Partnership for America’s Children is to engage, empower, and invest in a network of state and local child advocates who advance transformative, systemic, and equitable policy change for the well-being of every child across the nation. The Partnership has 48 member organizations in 39 states that advocate to improve policies for children at the local, state, and federal level. Collectively they represent over 90% of the nation’s children. Partnership members use Census data in their advocacy, and 56% of Partnership members are also KIDS COUNT grantees in their state, serving as that state’s data hub on children for policy makers, administrators, and nonprofits. The Partnership for America’s Children served as the national hub on the undercount of young children in the 2020 Decennial Census. In this role the Partnership formed and continues to co-lead a national working group of child-serving organizations that is working to improve the count of young children in all Census Bureau demographic products.
ZERO TO THREE is an early childhood development organization whose mission is to ensure that all infants and toddlers have a strong start in life. Staff at the organization regularly rely on Census to help policymakers understand the urgency of promoting policies that ensure strong early development. For the past five years, ZERO TO THREE has published the State of Babies Yearbook, which draws heavily on Census data sources to tell the story of babies and how experiences and outcomes vary by geography, race and ethnicity, and income, including data from every state, the District of Columbia, and Puerto Rico. ZERO TO THREE shares the concerns about the undercounting of young children and particularly children of color, who for several years have comprised a majority of children under age three. An inaccurate count obscures the true picture of young children who lack opportunities across states at a time of life when foundational development sets the stage for all that follows. This understanding of the power of data to identify and promote policies led ZERO TO THREE to become a partner in Count All Kids, to work with to improve the count of young children. ZERO TO THREE appreciates the Census Bureau’s efforts to improve the count of young children in the 2020 Census, bolstering this unique and powerful dataset, and its work toward continuing progress in the 2030 Census.
We submit these comments for the March 2025 Census Scientific Advisory Committee meeting.
We thank the Bureau for including a session on the ACS Research Program on Disclosure Avoidance Protections. This topic is of enormous importance to children. The disclosure avoidance approach used for reporting most of the 2020 Census data broke the relationship between children and the adults they lived with (although we understand that the relationship remains intact in the underlying data at the Bureau). If the same approach to reporting data is used for the American Community Survey (ACS) we will lose the single most important indicator of child well-being—child poverty data. The American Community Survey is our nation’s source of data on child poverty, deep child poverty, and children living in low-income households at the state or local level. State and local data on this topic is crucial for policy makers who are trying to determine where to make investments to improve the lives of children.
Dr. William O’Hare’s research revealed that the approach taken with the decennial census resulted in some highly improbable results. His research based on the Census Bureau applying the disclosure avoidance protocol used for the 2020 Census to the 2010 Census data shows that there were 163,077 blocks nationwide (1.5 percent of all blocks) that had population ages 0 to 17, but no population ages 18 or over, compared to 82 such blocks before differential privacy (DP) was applied. Discussions with the Bureau revealed that these improbable results were because the differential privacy method they used with the Census files sometimes separated data on children from data on the adults in the household.
More recently Dr. O’Hare and Dr. Ronald Prevost conducted a study that showed similarly improbable results using the 2020 Census product that provides block-level data, the Public Law 94-171 file, often called the redistricting file. That work is discussed in a blog by Dr. O’Hare and Dr. Prevost. Their work shows that there is significant geographic variation in where these “child only blocks” occur, and that White and American Indian/ Alaskan Native children are over-represented in child-only blocks, while Black and Hispanic children are underrepresented in them. They note that many users need accurate data at the block level, including “those working in political redistricting, school demographics, transportation analysis, or locating services such as fire stations and libraries”.
Child advocacy organizations have been very concerned about what would happen when a Differential Privacy method was applied to the American Community Survey (ACS), which provides the best source of child poverty data by state and locality. The ACS measures child poverty by looking at the income of the adults that kids live with, because that is what determines children’s access to resources. To tell whether children live in poverty you must look at the income of the adults in the household. If the Bureau applies the same differential privacy method to reporting ACS data, or any method that separates children’s data from adults, the ACS child poverty measures would be compromised.
While child poverty is the most important data element that would be eliminated or made significantly less accurate by the application of any privacy protection that separated children from adults in the data, there are other child well-being measures in the ACS that also would be damaged. These include data on family structure: how many children live with both parents, one parent, no parents, or with grandparents. Other important data that requires the children and the adults to be linked include: Parents lack secure employment; Household has high housing costs; Teens not in school and not working; Young child not in preschool; Child without health insurance; Children in single-parent families; Children in families where household head lacks a high school diploma; Children living in high-poverty areas. (These, with child poverty, are nine of the 16 data points chosen by the Annie E. Casey Foundation for inclusion in its KIDS COUNT data book.)
The American Community Survey data is very important in efforts to assess child well-being and improve policy affecting children. All our organizations use it extensively. A recent survey of child advocates also indicates the ACS is widely used among state advocates. The summary of that study (page 16) says, “The data analyzed here underscores the extent to which state and child advocacy organizations make extensive use of the data from the ACS in a variety of ways I doubt there is any other single source of data as important as the ACS for state and local child advocates or researchers.”
We appreciate that the Bureau also reported some data from the 2020 Census using a different privacy protocol that maintained the connection between children and adults. That data, in the Supplemental Demographic and Housing Characteristics (S-DHC) file is the only 2020 Census file which uses a variant of Differential Privacy (DP) that keeps the connection between children and their parents in the data processing step. The S-DHC file uses a variant of differential privacy called PHSafe. These tables are also known as the person-household join tables. Dr. O’Hare analyzed that data. He found that this file contained very little data. He concluded “Compared to the other DHC files the data from the S-DHC files are greatly restricted in terms of geographic levels, race/ethnic groups, and age breaks.
Recently, the Census Bureau announced a principled framework for disclosure avoidance they plan to use going forward. None of the 10 principles listed mention the goal of keeping children attached to their parents in data processing. This suggests that keeping children attached to their parents and producing quality measures of child well-being are not high on the Census Bureau’ priority. Principle 2 is “It Should Support Meaningful Assessment of the Impact of Data Protections on the Availability and Accuracy of the Statistics “. The discussion of the principle is entirely focused on degradation of accuracy. While this is certainly important, we believe that the Bureau should also consider the importance of making the data available. A key factor in this would be whether the data is needed for allocation of federal funding. Child poverty and income data is critical for policy making and for allocating many streams of federal funding. For that reason applying a disclosure avoidance approach that protects the availability and accuracy of that data should be a top priority. If the Bureau adopts a disclosure avoidance approach that involves a privacy “budget”, as formal privacy does, then the need for the data to be available and accurate to allocate federal funds should be a basis for spending more of the privacy budget on producing that data. (We note that the same concern applies to other data that is needed for allocation of federal funds.)
Therefore, we urge the Bureau to develop a privacy protection protocol for the American Community Survey that maintains the connection between adults and children, preserves the child poverty data and other essential data on children, and maximizes the availability of data for smaller geographies, smaller age groups, particular racial and ethnic groups, and other approaches to disaggregated data. No privacy protocol should be adopted if it eliminates or reduces the accuracy of child poverty data.
Thank you for considering these comments.
