Hogan et al. (2013) used Demographic Analysis to estimate that young children (children under the age of 5) in the 2010 Census had a net undercount of 4.6 percent. An interdivisional team at the Census Bureau is currently analyzing existing data sets from the 2010 Census to try to understand the reasons for this high level of coverage error. This report is part of a series of reports that examine the undercount of young children in the 2010 Census. It summarizes findings about households that may have erroneously excluded a young child when they completed their 2010 Census questionnaires.

The 2010 Census included a coverage improvement program to review the list of household members and identify potential coverage errors for follow-up. Three recently released reports summarized data from this Coverage Followup (CFU) operation. The first report (U.S. Census Bureau 2017a) looked at households that responded positively to one of the probes on the 2010 Census questionnaires about potentially omitted children. The report summarized the characteristics of households where the respondent was uncertain about including a child. The second report (U.S. Census Bureau 2017b) analyzed the characteristics of the young children added as a result of CFU and the characteristics of the households where they lived. Data from these two evaluations identified instances of potential coverage error involving young children. The third report (U.S. Census Bureau 2017c) analyzed the geographic distribution of each of these events—households with positive responses to a child-specific undercount probe and young children added during CFU. In that report, the authors recommended additional research to study the social and economic characteristics of the geographic areas with high positive-response rates and high CFU add rates. Segmentation group analysis was also proposed to understand if errors involving young children were largely concentrated in areas associated with hard-to-count characteristics.

In this report, we expand our analysis of the geographic distribution of the CFU results using two segmentation methods. These methods group neighborhoods with similar characteristics. The first method uses a marketing segmentation structure developed by Esri, a geographic information system company (www.esri.com). Esri created the Tapestry demographic and lifestyle segmentation for use in analyzing markets and consumers. The second method uses tract-level data from the Census Bureau’s Planning Database (PDB) to stratify the country based on specific social, economic, housing, and operational variables.

We use these two methods to identify areas with the greatest numbers and highest rates of positive responses to the child undercount probes—an indicator of the types of neighborhoods where respondents experienced some level of confusion about including young children on their census forms. In addition, we identify areas with the greatest numbers and proportions of their young children added as a result of the CFU operation. These results point us to neighborhoods where the CFU operation successfully added young children that the household respondent initially omitted in error. As we plan for the 2020 Census, it is useful to identify where respondents experienced challenges completing their 2010 Census forms. The Census Bureau could target these areas for special outreach and education efforts in 2020. The Esri Tapestry segmentation and the PDB provide us with a rich set of operational, demographic, housing, and socioeconomic data about the neighborhoods that experienced these enumeration challenges in 2010.