2020 Census: Lessons Learned and Best Practices for Researchers and Evaluators

Posted December 7, 2022

The year 2020 was, for many of us, a year of unforeseen and unexpected challenges. Unfortunately, the U.S. decennial census also experienced a challenging and unprecedented year in 2020. As they have done once every decade, since the U.S. Constitution established its requirement starting in 1790, enumerators attempted to count every person living within the U.S. borders. The Census is primarily used to redistrict U.S. House of Representative seats and adjust the Electoral College, but it also serves many other purposes – like determining where federal funds are distributed. Importantly, for the social scientists among us, the decennial Census has also been used as an important baseline for understanding key demographic information at every geographic level.

Counting all 330 million people in the U.S. is not an easy task in the best of times, and 2020 was not the best of times. The 2020 decennial census was faced with two major obstacles; the first of which was the COVID-19 pandemic. However, the Census faced another challenge on top of the pandemic—President Trump’s Administration abruptly cut the Census extension short—ending the follow-up efforts a month earlier than planned.

These factors negatively affected the Census’s ability to reach and represent several groups of people. Even in a typical year, the Census usually estimates undercounts or overcounts (due to imputation) across all racial and ethnic groups. Notably, in 2020, the Census estimated that it undercounted Hispanic/Latinx population in the U.S. by approximately 5% – in 2010, this estimate was 1.5%. The Census also estimated that it undercounted both Black/African American and Native American/Indigenous populations at higher rates than in past Census efforts, although not statistically significantly differently.[1] This all resulted in significantly lower response and less accurate data within the Hispanic/Latinx population.

So, what does this mean for our industry – and what lessons can we learn from the 2020 Census? There’s a lot to be extracted here, in terms of how researchers, like ILLUME, can do research that is representative and accurate, and improve how we think about reaching underserved populations and populations who may (for very good reasons) mistrust large institutions like utilities. We provide our recommendations below for researchers that use Census data regularly in their practice.

When Referencing the 2020 Census in Your Research:

  • If you need to use decennial Census data for weighting or sampling, make sure you understand the flaws in the 2020 data. Although we know the Census over and undercounted certain groups, this could vary widely at the state and tract level, so it’s not particularly feasible to adjust in a useful way based on their estimates without unknowingly introducing even more error. Using alternate data—like the 2015 – 2019 ACS survey estimates—might be a more reassuring option, even though the data are older. At a minimum, if you reference these data, caveat your research so readers will understand the implications.

Understanding Your Own Customers:

  • Collect your own data – know your own population. Program administrators have the power to understand exactly who they are reaching and not reaching through their programs. If you can, don’t be afraid to collect demographics from participants. Knowledge is power – and understanding who you aren’t reaching is your first step to understanding who you are underserving. 
  • Mixed mode survey efforts are your best bet for reaching diverse groups. Especially when doing general population research, choosing only web surveys can mean you miss key portions of the population. Believe it or not, people do still respond to mail surveys – there are many people in this country who have limited or no reliable internet access. Outreach letters that drive customers to a web survey are also a great way to reach people. Don’t forget to consider translating your survey—and your outreach materials—into commonly-spoken non-English languages in your region.
  • Do what you can to highlight the legitimacy of your research and build trust in it. Prepare your call center for questions. If you can, add a landing page on your website that has the details of the research you’re doing. Offering incentives as a thank you for customer time is a great way to show customers you are serious about their input. Lastly, letting customers know that their information will be kept confidential and not shared outside of the research team is critical in trust-building.
  • Census and demographic data only tell you so much – what do your communities look like, and what do they need? ILLUME has done some more granular research to better understand where pockets of customers exist that might be missed by utility program offerings, and more importantly – why. For example, we examined the needs of the Vietnamese and Korean-speaking communities in a large metro area, and the results of our study led the utility to translate point of purchase materials at stores in very specific areas. There isn’t a broad-brush solution to understanding underserved communities; each one is different and unique and needs to be examined that way.

We hope 2030 is going to feel very differently than 2020 did. In the meantime, doing your own research that takes learnings from the Census—both good and bad—is the best way to keep your finger on the pulse of your customers and their needs.  

[1] 2020 Census Undercounted Hispanic, Black and Native American Residents – The New York Times (nytimes.com)