Read more articles by Betsy Gardner

By Betsy Gardner • July 8, 2019

Do you count? Most people would make the assumption that they do, on some level or to someone. But it’s possible that the federal government has a different answer than your friends and family; you might not count to it at all. Consider the following questions: Do you live without reliable internet? You may be hard to count in the 2020 census, the first offering online responses. Have a child under five years old? Your child is considered hard to count, with 2.2 million children under five missing from the 2010 census. Rent your home? You may be hard to count, as homeowners are far more likely to return census questionnaires than renters. Identify as Hispanic? You may be hard to count, due to possible language barriers or concerns around ethnic profiling.  


Hard to count is an official categorization from the Census Bureau and there’s a working group dedicated to this topic. Although not an exhaustive list, communities that are considered hard to count by the federal government are “racial and ethnic minorities, persons who do not speak English fluently, lower income persons, homeless persons, undocumented immigrants, young mobile persons, children, persons who are angry at and/or distrust the government, and LGBTQ persons.” Individuals in these groups are most likely to have a low self-response rate, which is what makes them hard to count. The Census Bureau sends enumerators to talk with every non-responding household individually. This follow-up is time-consuming, difficult, and costly. But not counting everyone has its own costs too. 


According to a study of the 2010 census from the George Washington University, the median state loss per uncounted person was $1,091. And when people aren’t counted in the census, it delivers a fairly clear message about where priorities lie and what types of communities are deserving of attention, funding, and political representation. These discrepancies between the counted and undercounted advantage and disadvantage different groups, and often reinforce existing discriminations. So how can local governments and community organizations try to shift this balance, and ensure an inclusive census in 2020? Thankfully, there’s a map for that. 


“We hope everyone is concerned about the census,” said Steven Romalewski, Director of the Mapping Service at the City University of New York’s Graduate Center who led the team that created the Hard to Count Map. Community organizers, journalists, teachers, and mayors have all praised the interactive Hard to Count Map. In 2009 they made a version of the same map at the request of a coalition of philanthropic foundations and civil rights groups. For 2020 a similar constellation of organizations, in particular the Leadership Conference on Human and Civil Rights (LCHCR), came together a couple years ago to start the same work for the upcoming census. CUNY’s 2020 Hard to Count Map was launched in October 2017.


The Census Bureau also published a similar map in 2010, calculating and publishing a hard to count score for census tracts. However, for 2020 it elected to change the data, and utilize a predictive statistical model that would try to identify census tracts that would likely have low self-response rates. Romalewski, working closely with several census experts, was concerned that a predictive model of low response scores (LRS) would not be reliably predictive enough for organizations and local governments that needed to plan their outreach. Instead, his team decided to use the actual self-response rate from 2010 as an initial indicator of which tracts would be considered hard to count. 


In 2010 just over half of the households in Tract 0387.00 in  Kings County N.Y. returned their census questionnaires. This tract is one of the hardest to count in the country, requiring expensive and difficult in-person follow up by the Census Bureau. Tracts with 2010 mail return rates of 73% or less (in the bottom 20 percent of return rates nationwide) are shaded on the map 



Romalewski did include some of the features from the Census Bureau’s LRS map, as it showed recent demographic data with implications for outreach and response rates. By including 2010 historical data and demographics from the most recent American Community Survey, the CUNY 2020 Hard to Count Map lets advocates and governments decide what metrics they’ll use for targeting their outreach and support. The HTC map from Romalewski’s team also provides hard-to-count population data for other areas, not only  census tracts. Users of CUNY’s map can easily select areas such as congressional districts and counties, especially important if they’re working in local or municipal government and need to review certain geographic limits. 


Demographic information for Lowndes County, A.L. Almost a quarter of households in Lowndes County are single parent; Census Bureau data shows that single parent households have lower self-response rates.  


Local and regional governments have been very enthusiastic about this map. On Twitter, mayors and public health commissions have reached out to Romalewski to thank his team for their work. Officials understand the importance of counting everyone, at both a human level and an administrative level. Telling constituents that they count is a basic validation of their existence and their importance to the greater community. Historically under counted and marginalized communities receive less resources and decreased representation, which has cascading effects on things like the quality of schools, healthcare and infrastructure. The Census Bureau discusses how the census impacts communities, giving examples that include affordable housing for the elderly and emergency preparedness; to leave out a segment of the population because they have historically been harder to count denies them those benefits. Counting everyone is an important step in creating inclusive and representative policies.  

Tweet from Mayor supporting the HTC map
Tweet from Mayor Daniel Corona of West Wendover, N.V. demonstrating how local government can use the Hard to Count Map in their census planning process.


There is also a very clear data benefit. The National League of Cities (NLC) highlighted CUNY’s map in an article urging cities and towns to establish their outreach, budgets, and community engagement partners. As the NLC pointed out, collecting census information provides “meaningful data for municipal operations.” Local chief data officers and chief innovation officers should be especially involved in the census process, both because of the new digital responses in 2020 and because they will rely on data coming from the census. It’s in their best interest to have clean, quality, and comprehensive information. 


Finally, there are real fiscal consequences for cities, states, and counties that undercount their population. In addition to the 2010 GWU study which places the median state loss per uncounted person at $1,091, there are specific programs that are at risk when hard to count populations are missed. The Mid-Atlantic Regional Council, covering the Kansas City metropolitan area, had a presentation with the KC Census Funders Partnership that showed exactly how much the region missed due to undercounting in 2010, and forecast the loss for a similar undercount in 2020. The affected programs include Women, Infants and Children (WIC), Medicaid, SNAP, and highway planning and construction.  



Slide from the Mid-Atlantic Regional Council presentation breaking down potential losses by county.



Another benefit of CUNY’s interactive mapping application is the flexibility for users, who might want to filter other information that isn’t included in the Census Bureau’s map. For example, Romalewski’s team incorporated the location of public library branches on the map, in partnership with the American Library Association. This is important because libraries have free internet access and 2020 is the first time responders can submit the census questionnaire online. Romalewski said that his team did an analysis “that showed that 99% of public libraries are in or within close proximity to hard to count tracts, or tracts that had low internet penetration.” This is crucial knowledge for anyone looking to assist hard to count populations, especially if local governments or community organizations want to create a hub for digital responses.    


Libraries overlaid on a section of Appalachia. As noted in the left-hand column, 34% of households in Tract 9571.00 in Mingo County W.V. either have no home internet subscription or only dial up. 



In addition to the robust data, CUNY’s map also centers on the user experience. The CUNY team focused on making the map as easy to use and intuitive as possible so anyone can go to it, start clicking around, and instantly get information. They made sure to provide narrative around the numbers as well, so users can quickly place the data in context. Almost everything a user might want to click on is clickable; there are numerous links leading to further fact sheets, original data sources, and even a list of groups doing census outreach in each state. Romalewski knew that without that kind of context and background it can be hard to efficiently understand and use the data. Those involved with the census count appreciate the additional context, as it helps them focus limited time and funds to the appropriate targets. 


More updates and enhancements are in the works for the Hard to Count Map, and once the 2020 census is underway Romalewski’s team will be integrating daily data to give cities, counties, and community organizations the opportunity to further tailor their outreach. During the actual 2020 enumeration the Census Bureau has offered to publish daily the share of households that have responded, based on the mailings and online submissions, at the tract level. That will be incorporated into the CUNY map so users will be able to see which areas are doing well at responding to the census, and which are not. Romalewski envisions public officials and local organizations using that data “to decide if they need to redirect their resources into certain areas and not to others” and have an agile response based on real-time data.  


Romalewski sees his map as the most “holistic way of getting the information out there and making it operational” so that local and regional groups can make the best data-driven decisions about census collection for their communities. The efficiency and agility enabled by the Hard to Count Map will guide public officials and advocacy groups to ensure the inclusion of traditionally marginalized communities, and send the message: you count.