The Digital Divide and Census 2020

This is the eleventh blog post in an in-depth series exploring the history and future of the US Census. If you’re interested in other subjects related to Census 2020, check out the list of all our blog posts about it at the end of this post.
The Decennial Census in 2020 will be conducted primarily online. People have concerns about the new format such as: Will the Census Bureau be able to create and disseminate a web application in time for the 2020 Count Day? And will response rates fall in areas with less access to Internet services? Areas affected by the digital divide tend to already be Hard-to-Count populations such as group housing units in cities and poorer populations.  These questions have serious impacts because a national undercount would distort the national population figure; at state and local levels, an undercount can lead to less federal aid being sent to the state and disproportionate representation in Congress. It is important to modernize the census data collection process, but it is also important to consider the voices that may be left out. This blog will discuss why people have not been counted by the census in the past, how the Census Bureau intends to use internet collection in 2020, the state of internet access in the City of Detroit, and how this may affect the population count for the city.

Overcounts and Undercounts

After the 2010 Census, the Census Bureau suggested the census had produced a net overcount of 0.01%, or about 36,000 people. This was an improvement on the 2000 Census, with a net overcount of 0.49%, and on the 1990 Census, with a net undercount of 1.61%.

How can the census possibly count people who aren’t really there?

The Census Bureau estimated that duplicates of individuals comprise most of the overcounted population (84.9%). Duplicates usually occur when parents include a child who is away at college on their census form and the child fills out a census form at school, when people move and are counted in both locations, and people with two residences that fill out a census form at both.  The rest of the overcount can be attributed to people who died after the census was taken, and those who were born after it.

However, the 0.01% overcount number hides an undercount of certain populations; the overcount is simply the numerical aggregate of estimates of error. The Census Bureau estimates that renters were undercounted by 1.1%, and African Americans were undercounted by 2.1%. The reasons for undercounting vary, but an undercount is the result of non-participation.  Most groups of Americans who are undercounted are considered to be Hard-to-Count populations and special outreach efforts are developed by the Census Bureau to engage them in the census count.

Hard-to-Count populations are those groups of people that have historically had lower census response rates, and include minorities, those who live in poverty, the homeless, immigrants, those whose primary language is one other than English, and young people (For more on Hard-to-Count populations, check out our previous blog). These factors make it significantly harder for census counts to be accurate in urban areas, especially those with populations in poverty, like Detroit.

Detroit’s Participation Rate, 2010

So how did Detroit fare in terms of census participation in 2010? Not well.

One way the Census Bureau can evaluate participation in Hard-to-Count tracts is by measuring the mail-in participation rates of various tracts. In the typical census year, a majority of census information is compiled through the use of mail-in forms sent to address lists compiled during the Local Update of Census Addresses. Most people fill out the form and return it, but those who don’t have to be followed up on by canvassers. The Bureau keeps track of the number of people in a given census tract who return the mail-in form, and divide it by the total number of people living in the tract to create a Census Participation Rate.

Detroit’s Participation Rate fell from 70% in 2000 to 64% in 2010. Some census tracts in the city fared much worse, with one tract returning just 14% of mail-in forms, and many others hovering around 50%, significantly below the state’s 78% participation rate. Much of this can be attributed to the reasons for undercounting discussed above: residents living under the poverty level, minority populations, and people whose native language is not English.

Interact with the CensusExplorer Map.

Overall, the State of Michigan ranked 5th in the nation in Census Participation Rate, at 78%. Livonia, MI was the municipality (with more than 50,000 residents) with the highest response rate nationally (88%). The disparity in participation rates between Detroit and the Metro suburbs can also lead to disproportional allocations of federal and state funds. In other words, a less accurate count in areas that need more assistance can lead to less help being sent.

Taking the Census Online

As part of a plan to minimize cost increases from the 2010 Census (Check out our blog on the Census’ Budget!), the 2020 Census aims to collect a majority of its information from an online portal. This changes the entire approach of census collection efforts.

Instead of first receiving a mail-in form and having a canvasser interview you if you do not return it, the 2020 Census will send a letter to each address in the Bureau’s database with a personalized PIN, which will authorize access to a web application to input demographic information. If responses are not recorded for an address that was sent a PIN, canvassers will then visit the address. It is still unclear whether or not a mail-in form will accompany the letter with the PIN or if it will need to be requested, but mail-in responses are only expected to comprise of about 15% of recorded information. The hope for the Census Bureau is that it can hire fewer canvassers by improving the routes that canvassers travel between houses and decrease the need for canvassers in the first place.

The major concern with this change, however, is that areas with lower-than-average response rates also have fewer households with internet access than average. For these census tracts, it is going to be challenging to gather information from people without sending in a canvasser, and participation rates will presumably fall.

Let’s consider how the change to the online format in the 2020 Census may affect Detroit by looking at broadband coverage for every census tract in the city. Tracts have been analyzed for access to both 3 and 10 Megabits per second (download speeds) broadband connections based off of Federal Communications Commission data. Tracts that had less than 20% of households with a connection of either speed were identified as areas with “Low Broadband Coverage”. Comparing the 53 tracts that were labeled as “Low Broadband” with 2010 Census Participation Rates, it was found that those tracts had a mean mail-in participation score of 58.1%, and a median of 58%.

This means that on average, tracts with less access to the Internet do not turn in their mail-in census forms at as high of a rate as the rest of the city. Interestingly, the Mail-In Participation Rate for the city increases to 72.2% without these 53 tracts. This should be an early-warning sign for the city, as the online census will more than likely lower participation rates further in these 53 tracts, and the need for extra canvassing in this area is probable.

The 2020 Census poses a number of problems that are going to require innovative solutions. Communities with less access to broadband may find it more difficult to complete a census that is going to be done primarily online, but there are several solutions that may work if done together. It is possible that the Census Bureau changes their approach to the mail-in form, and ends up sending it to every household that receives a letter with a PIN. The Bureau would then allow respondents to send back the mail-in form if they did not have access to the online portal, which may increase participation rates.

There has also been discussion that the Census Bureau may partner with community centers to provide access to computer terminals to allow those without internet access to participate in the census. This would also result in an increase in participation, as people would not need to mail back the form the Bureau sent them, but simply fill out the census online near their homes. Community organizations could also assist in raising awareness for the upcoming census, and making sure to target these neighborhoods. Residents of neighborhoods with low Participation Rates in 2010 should also be targeted for outreach, to ensure that they complete their form and that canvassers know where to look for people.

Census Participation Rates have been falling for decades (see maps below for reference) and although it is unlikely that the trend will be bucked in 2020, we should do the best we can to maximize response rates, including addressing issues of digital divide that create challenges to the inclusivity of the new online format.