It shouldn’t be surprising that the Census Bureau does not perform with 100 percent accuracy when carrying out its official Census every 10 years. Some people are difficult to reach, and others may decide not to fill out the form. If the number of people who do not fill out the form is small and randomly distributed this wouldn’t be a large problem, however, if this population is not random then there is a significant risk that some areas will be disproportionately undercounted. Throughout the history of performing Census counts, researchers have learned a lot of information about populations that are typically undercounted, known as Hard to Count Populations (HTC).
Which Groups are Traditionally Hard to Count?
Researchers at the Census Bureau have identified a list of populations that are traditionally harder to count than others. These include:
- Minorities (especially African Americans)
- People in poverty
- People who are homeless or living in non-traditional homes
- Immigrants (especially illegal)
- People who primarily speak a different language than English
Additionally, many of these populations exhibit a higher than average distrust or fear of the government due to perceived inequalities in treatment, and therefore make a conscious effort not to fill out a Census form. For example some immigrants may believe the information collected by government agencies may be used in deportation efforts or had adverse experience with government in their home country, or people in minority groups may distrust government agencies due to perceptions that other government units such as police are unjust.
The HTC Score The Planning Database’s Low Response Score (LRS)
The Census publishes the Planning Database annually by assembling a number of indicators from the Census, American Community Survey, and operational data (like the Census Mail Return Rate). These are used to calculate a Low Response Score, which measures the perceived difficulty of counting both block groups and tract levels based on mailed-in self-response. There are 25 LRS predictor variables, with percentage of renters in a block group, percentage of people aged 18-24, and percentage of households headed by unmarried females as the highest predictors of low response rates.
The Low Response Score is negatively associated with the 2010 Census Mail Return Rate. These data can be used to guide the Census Bureau’s activity leading up to Census 2020 to ensure outreach is targeted in areas with high levels of hard to count populations. However, the current Planning Database’s LRS does not account for the use of Internet as the primary mode of self-response in Census 2020. This means that different segments of the population will respond in various ways. Consequently, the LRS is score not as accurate as it could be and characterizes some geographic areas as harder to count than they really are.
In the past, the Census Bureau has used a database that contains an “HTC score” that operates in the same way as the LRS by indicating the perceived difficulty of counting Census tracts. The scores are assigned using the last available Census and are merely estimates and therefore cannot be used to determine actual levels of undercount.
The Census Project also created an interactive map of hard-to-count tracts and can be zoomed into specific Congressional and state legislative districts. It includes data such as the 2010 Census Mail Return Rate and demographic information about groups that are generally hard to count, such as racial and ethnic minorities, persons living in multi-family housing units, and low-income populations.
Why Worry About Hard to Count Populations?
There are many reasons why it is important to make sure that the Census gets an accurate count. The Census plays a major role on a federal, state, and local level.
The original justification for the Census in the U.S. constitution is to fairly allocate the 435 seats in the House of Representatives according to population. This means that if a state’s Census count drops, or doesn’t rise as fast as other states they could have less representation in Washington. This was the case in the 2010 Census where Michigan was the only state to have a population loss in the Census, and one of the few states to lose seats in the House.
In addition to representation at the federal level, the state of Michigan also uses Census counts to draw its congressional districts, with the requirement that districts be roughly equal in population. This means that if an area of the state is undercounted by a large enough margin, they may receive less representation at the state level than they should.
The Census count affects more than $400 billion in the allocation of federal funds, a lot of which is directed at aiding populations in need. This can be particularly alarming because HTC populations are often those who could benefit most from receiving federal aid. If an area’s HTC populations are not well counted, they will be receiving a disproportionately low amount of aid, which could have lasting negative effects.
Census counts directly affect how states distribute funds on many fronts. Due to the complicated methods that governments use to allocate funds, it is difficult to put a dollar value on how much funding is lost with each decrease in population. However, many sources of funding are still strongly tied to relative population levels, so the issue remains pressing. A recent study at George Washington University finds Michigan receives about $1,400 per person counted in the US Census. Similarly, during the 2010 Census, Detroit’s population fell below the 750,000 mark which made it ineligible for special tax rates; this put the state in danger of losing significant funding. The Michigan legislature went on to lower the population threshold to 600,000, but saves like this may not always be an option.
How Detroit Stacks Up
Detroit has a large number of HTC populations relative to the rest of the country; this is due to its somewhat large population of minorities, poor, homeless and foreign language speakers. This makes the city vulnerable to potentially large disparities in funding or representation, which means increased and focused efforts would need to be made to count HTC populations in order to correct inequalities.