Responses to the Census Bureau’s Demographic and Housing Characteristics Demonstration Product Raise Many Data Issues Related to Children

Dr. Bill O'Hare

The next data product that will be released from the 2020 Census is the Demographic and Housing Characteristics (DHC) file which is scheduled to be released in May 2023.  This file is particularly important for child advocates because it is the first file from the 2020 Census that will provide data for young children (ages birth to 4) as well as data by single year of age for all children.

 The U.S. Census Bureau is using a new method called differential privacy (DP) to help protect confidentiality and privacy of Census respondents in releasing data from the 2020 Census.[1]   Like previous data from the 2020 Census, the DHC file will have differential privacy applied.

In March 2022, the Census Bureau (2022a) released a DHC Demonstration Product with DP applied to the 2010 Census tables that will be in the DHC file and asked data users to analyze the data and provide the Census Bureau with feedback.  The feedback from this Demonstration Product will be used to improve the accuracy of data in a second Demonstration Product that will be issued prior to finalizing the parameters for the DHC File to be issued in May 2023.  Applying DP to the 2010 Census data allows researchers to compare the 2010 Census data with and without DP to get an idea of what the impact is likely to be on 2020 Census data.    

The Count All Kids campaign provided the Census Bureau with a report that focused on the population ages 0 to 4 with respect to Unified School Districts and Places (O’Hare 2022 ) and the Partnership for America’s Children also submitted comments.[2]  Unified School Districts and Places were examined because they have policy-making responsibility for programs focused on children.

Recently the Census Bureau released a summary of the feedback they received on the March 2022 DHC Demonstration Product  Their summary focused on 14 submissions that related to the DHC Demonstration Product and particularly the ones related to accuracy of the Demonstration Product data.

 Concerns for data on children were a prominent topic in the submissions. According to the Census Bureau’s review of submissions,  “The most common subtopic was age data on children.”  When the U.S. Census Bureau (2022b. Table 3) cataloged the topics and subtopics addressed in the submissions, they noted that 3 of the 12 submissions focused on children, 2 of the 12 focused on age 0 to 4, and 2 focused on school age children.   One other subtopic mentioned in the submissions was household type by presence of children. The review of submissions by the Census Bureau (2022b. Table 5) found the most common recommendation was increased accuracy for age.  The Census Bureau report concluded, “Accuracy for what were considered key single years of age for children and school districts was a  specific age-related concern.”  

The Census Bureau also noted the concern about data for Unified School Districts was mentioned often in the submissions, and this is obviously related to children.  At the National Academy of Sciences workshop on Differential Privacy held June 21-22, 2022, the Census Bureau noted the concerns with Unified School District data and said they planned to look into ways to improve the data for children in Unified School Districts.

It is gratifying to see the Census Bureau note stakeholders input related to data for children and this will hopefully result in more accurate 2030 Census data for children. 

[1] The terminology in this arena can be confusing.  Differential Privacy is sometimes called “formal privacy.”  The system developed for the 2020 Census DHC file has also been called the Top-Down Algorithm or TDA. Since the application of differential privacy occurs within the Census Bureau’s Disclosure Avoidance Systems (DAS) that term has sometimes been used to describe the use of Differential Privacy. To avoid confusion, I use the term differential privacy (DP) here to distinguish the version of DAS that includes DP from other versions of DAS.

[2] There are about 11,000 Unified School Districts and most of them include multiple schools within the district. There are a little more than 29,000 Places including over 19,000 Incorporated Places and nearly 10,000 Census Designated Places (CDPs).

 

References