This post has been updated.
Chris Dick, a former Census Bureau statistician and branch chief, is head of the government practice at the data science firm Civis Analytics.
This morning, the Supreme Court blocked the Trump administration’s efforts to add a question about citizenship to the 2020 Census — at least, for now. This is a huge win for democracy: There is plenty of compelling research on why the citizenship question would depress response rates, particularly among immigrant, African American and Hispanic groups. But the danger isn’t over: Declining trust in government and a lower likelihood that people will respond to surveys mean there is still a risk of nonresponse between 20 and 30 percent.
The 2020 Census is on the Government Accountability Office’s list of high-risk projects, and the explanation the agency offered for that assessment doesn’t even mention the citizenship question. The census isn’t just a matter of government record-keeping. If we get this wrong, it will have repercussions for years to come.
The decennial census guides the allocation of $800 billion in federal funding, determines congressional representation and districting, and underpins important business decisions. Undercounted areas will be underfunded and underrepresented in Congress, and our population data will be inherently flawed. And the undercount can’t be rectified for a decade.
Starting mid-March 2020, the Census Bureau will mail forms to everyone in the United States. Every member of every household needs to respond and can do so online, through the mail or over the phone. If a household doesn’t respond, the Census Bureau will send a so-called “enumerator” to go out and knock on every door up to three times.
This is where things get both complicated and expensive. The Census Bureau needs to hire approximately 500,000 temporary workers to serve as enumerators, which is particularly difficult given the low unemployment rate. Doing so costs taxpayers $55 million for each percentage point of households requiring follow-up. If those three tries are unsuccessful, the Census Bureau will ask a neighbor for information or use existing government data sources to make an educated guess about the household.
There are always “hard-to-count” populations, each with a unique motivation for nonresponse. If the Commerce Department manages to come up with a more compelling rationale for the citizenship question and it ends up on the form, that query would add an element of fear for immigrants and other minority groups that their answers could be used against them. Even if it’s never added, the buzz around the question has already cast a pall over the census.
To achieve this monumental task, we need to treat the 2020 Census as if it’s the country’s most important marketing campaign — one we’re all responsible for.
Fortunately, organizations focused on the census are taking the upcoming count extremely seriously — and they aren’t new to the task of getting their constituents to respond. Tribal, state and local governments partner with community organizations to create Complete Count Committees and to serve as field offices for get-out-the-count campaigns. National nonprofits such as the National Association of Latino Elected and Appointed Officials, Census Counts and the Funders Census Initiative, which distributed $40 million in grants to power response efforts for the 2010 Census, are supporting these efforts to give all communities a fighting chance for fair representation.
Commercial and political organizations know that data-driven, one-to-one marketing is key to success, and this is no different. You don’t just need the right message: You also need it to be delivered by the right person, at the right time.
That starts with identifying each hard-to-count population and what motivates them in as much detail as possible. Census Bureau research is a good starting point, but other resources and organizations can provide greater detail — for example, the Graduate Center at the City University of New York has a helpful map of high-risk regions with regularly updated details down to the neighborhood level.
This process will inform messaging development, but that too needs to be tested, because gut instinct about what will work is often wrong and certain messages can backfire. For example, research conducted by my organization, which consults on data science issues, found that emphasizing the strength of census data privacy and security practices actually made most groups less likely to respond.
Cadence matters, as well. In the Houston area, city and county governments, together with 50-plus local organizations, created Houston in Action to ensure that every hard-to-count group is contacted with the right message but not inundated with too many or conflicting communications.
Who delivers the message is just as important as the message itself. Consumer brands are focused on influencer marketing because they recognize that certain messengers are more persuasive for specific audiences. The same thing applies to the census, though of course, the “influencers” vary widely by community. In many cases, a trusted community figure (for example, a well-known pastor) is far more effective than a group of canvassers from outside the community. The key is locating and engaging those individuals for each and every district.
All this sounds like a lot of work, and it is. It won’t be easy. But with the right approach, the right tools and the right people, we have a fighting chance. The people of the United States — all of them — deserve our best efforts.