Map Sets for New York Counts Partners

Maps have proven to be a powerful tool for organizing efforts to have a fair and accurate 2020 Census. One of the goals of the New York Counts coalition is to centralize as many resources as possible for New Yorkers, for each of our partners, and for all who are concerned about seeing that everyone is counted this year.

CUNY Graduate Center Census 2020 Hard to Count map

What is the “hard to count” population and why does it matter?

The goal of the decennial census is to count each person in the United States based on their residence as of April 1. For the 2020 census, each household in the U.S. will either receive mailed instructions on how to fill out the census questionnaire online, or they will receive the actual questionnaire. The Census Bureau asks that as many households as possible submit their responses to this questionnaire via the Internet or by mail — this is the self-response component of the decennial census.

In prior censuses, the self-response rate in many parts of the country has been relatively high. But in other parts of the country and for some population groups more than others, the self-response rate has been relatively low. Households may not have submitted their census questionnaire for various reasons, such as having language difficulties, concerns about trust in government, or otherwise.

These areas and population groups are considered “hard to count”, because the Census Bureau sends enumerators into the field to talk with each non-responding household one-by-one. This “non-response follow-up” component of the census can be difficult, time-consuming, & costly (to the Bureau, and to taxpayers). And if these groups and their communities are not counted fairly & accurately, they will be deprived of equal political representation and vital public and private resources.

The goal of this map is to highlight the areas of the country that are hardest to count, and to provide information to local, regional/statewide, & national organizations who are working to make sure these hard-to-count areas & populations are fully counted to help ensure a fair and accurate census.

Defining hard-to-count (HTC)

For the purpose of this map, a census tract is considered hard-to-count (HTC) if its self-response rate in the 2010 decennial census was 73% or less. If 73% or fewer of the tract’s households that received a census questionnaire mailed it back to the Census Bureau, it is shaded in light orange-to-dark red as a hard-to-count tract on the map.

This measure of self-response for the 2010 census is called the mail return rate [PDF]. It represents the percent of occupied housing units only whose residents answered the census in the self-response stage of the count.

The 73% mail return rate threshold is used because it represents all tracts nationwide that are in the bottom 20 percent of 2010 mail return rates — i.e., the worst 20% of return rates. This is consistent with the definition of hard-to-count tracts from the 2010 census outreach campaign.

Some tracts do not have mail return rates. During the 2010 census, households in approximately 400 tracts did not receive a census questionnaire. Instead, these households were counted by the Census Bureau using solely in-person enumeration as part of the Update/Enumerate [PDF] program. These tracts were either very rural, did not have traditional street addresses, or both (often located on Tribal Lands, or in rural parts of Alaska or Wisconsin, and in some vacation areas with high seasonal vacancies). Although they do not have mail return rates, these tracts nonetheless are considered hard-to-count because of the cost & difficulty of in-person enumeration.

Other HTC metrics

Other ways of identifying and describing hard-to-count populations include:

  • Low Response Scores
  • The Census Bureau has developed a statistical model that uses population data to assign a “low response score” to each tract. The Bureau states that these scores “predict low census mail return rates and are highly correlated (negatively) with census and survey participation.”The Bureau’s research [PDF] indicates that this statistical model explains only 55% of the variation around the predicted response rate. The Bureau will be refining this statistical model leading up to the 2020 Census using more recent demographic data. We have incorporated these scores into the Census 2020 Hard to Count map in order to provide another metric for evaluating hard-to-count areas (see this update).
  • Population groups with increased risk of being undercounted
  • Historically, the census has undercounted young children, people of color, rural residents, & low-income households at higher rates than other population groups. (See Race and Ethnicity in the 2020 Census [PDF].) Also, groups with low self-response rates in prior censuses or census tests include “linguistically isolated” households; frequent movers; foreign born residents; households below the poverty line; large (i.e. overcrowded) households; low educational attainment households; & single-parent headed households. And people who distrust government authorities and/or have been or could be targets of law enforcement or heightened surveillance may be less likely to respond to the census. In the Census 2020 HTC application, statistics on these groups for each tract are presented when a tract is selected on the map.
  • Households with no computer or inadequate Internet access
  • The Census Bureau plans to encourage most households to answer their 2020 census questionnaire via the internet. As a result, households with poor internet connectivity or, worse, no computer will be at risk of being undercounted. The Census 2020 HTC map application highlights tract-level household internet access based on data from the latest American Community Survey estimates (see this update).
Data sources

In order to map hard-to-count tracts and population estimates related to them, we used several data sources starting with data from the U.S. Census Bureau itself:

  • The Planning Database, which provides operational variables from the 2010 census for each tract (especially the mail return rates) not found in any other publicly accessible database.
  • The American Community Survey (ACS), providing tract-level population estimates for the 2014-2018 period. ACS estimates were downloaded from the Census Bureau’s data.census.gov online application. Earlier ACS estimates for the 2013-2017 period had been downloaded from the Census Bureau’s American FactFinder application and from the National Historic GIS website via IPUMS NHGIS, University of Minnesota.

The map uses tract boundaries & Congressional and State Legislative district boundaries from the Census Bureau’s TIGER databases. The congressional districts correspond to the 116th Congress. The state legislative boundaries initially represented district lines as of 2016, but have been updated to reflect changes since then in North Carolina and Alabama (current as of 2019).

In order to calculate district-level population estimates for people living in HTC tracts, we used the Missouri Census Data Center’s “Geographic Correspondence Engine” to assign tracts to districts. If an HTC tract was split across more than one district, we used the Geographic Correspondence Engine to determine which district had the greatest share of the tract’s population, and allocated that tract’s population entirely to that overlapping district.

For each Congressional district and state legislative district, the Census 2020 HTC application provides the district representative’s name and contact information. This is accessed through ProPublica’s Congress API & the the Open States API.

Credits
  • The Census 2020 HTC map application was developed by the CUNY Mapping Service at the City University of New York’s Graduate Center. The Mapping Service, part of the Center for Urban Research, engages with foundations, government agencies, businesses, nonprofits, and other CUNY researchers to use spatial information and analysis techniques to develop and execute applied research projects. The Census 2020 HTC map reprises a similar application developed by CUNY for the 2010 census.
  • The Census 2020 HTC mapping effort was coordinated by Steven Romalewski. The Mapping Service’s application architect David Burgoon designed, created, & coded the Census 2020 HTC application. Will Field, our senior application developer, is updating & enhancing the site through 2020 and beyond. Simon Liu, a graduate student at Lehman College’s Geographic Information Science program, initially analyzed much of the data behind the map. Valerie Bauer, also a Lehman College graduate student, is helping with the data & maps. The Census 2020 HTC website relies on the Graduate Center’s server infrastructure and the support of the Graduate Center’s information technology team.
  • Funding support for the project is provided by the 2020 Census Project (special thanks to the Bauman Foundation).
  • The Mapping Service worked closely with The Leadership Conference and census consultants Bill O’Hare and Terri Ann Lowenthal, the Georgetown University Law Center on Poverty and Inequality, and other civil rights groups to seek guidance and feedback on the mapping application. Logic Department also provided helpful user interaction feedback and advice.
  • The application itself is built using a mix of open source and proprietary tools and technologies including:
    • MapBox, which is providing generous support hosting nationwide tract-level map layers that are featured on the HTC map;
    • Esri’s ArcGIS software, including map layers hosted via ArcGIS Online;
    • Carto’s basemap for reference;
    • Google’s places API;
    • Leaflet and Esri’s Leaflet tools;
    • Bootstrap, Vue.js, Lodash, jQuery, and Visual Studio.

New York 2020 Census Interactive Map by Hester Street

More Civic Asset Mapping Tools from Hester Street

New York’s Complete Count Committees

More About the Complete Count Committee Map