We are excited to share with you the first research we have seen that documents that state outreach activities in 2020 did improve census participation, particularly in communities where young children are at high risk of being missed. This is excellent news. It is particularly important since the Census Bureau has not yet begun to develop its communications campaign for 2030 and is well behind the process it followed in 2020. This makes it even more important for states to conduct outreach efforts.
Specifically, research by Phil Rocco, Chair and Professor of Political Science at Marquette University, showed that household self-response improved in Census tracts where states made any investment in Census outreach, and that this increase was particularly high for communities with high numbers of families of color, high numbers of people with linguistic barriers, and high numbers of renters. Each of these factors correlates with an undercount of young children.
We have been saying for a while that child advocates and child serving organizations should be working now to persuade their states and localities to form Complete Count Committees to conduct Census outreach in 2030 and encourage those committees to include in their work a focus on the count of young children. This research should help you show funders and policy makers that creating these Complete Count Committees will indeed help the count of young children.
This is new research that has been submitted for publication in a peer-reviewed journal. Professor Rocco is also the author of a book, Counting Like a State, which focuses on what state and local governments can do to improve the quality of census data. As you consider how to build on the research we are sharing here, you may find it helpful to get his book, available here. (This discount code is effective until October 8, 2026: 24COUNTING26.) Chapter four discusses what states undertook for outreach activities, documents the funding allocations by state, and identifies some of the factors that made states more likely to invest in Census outreach.
For this research, Professor Rocco looked at whether state investments in census outreach improved self-response rates. Self-response refers to households that received mailings from the Bureau asking them to fill out the census and did so, either online, by phone, or by returning a form. Two-thirds of all households did so. The remaining households were counted by an enumerator (“census taker”) going to their residence, with administrative data, by getting information from others who know the household, or from imputation.
It is important that state outreach improved self-response because Census Bureau research has repeatedly shown that self-response is the most accurate way to count a household, including in the 2020 Census. (See table 9 in the appendix.)
To conduct this analysis, Professor Rocco compared the self-response rates in census tracts where state outreach investments were made with similar census tracts were not made. The data on census tract response rates came from the Census Bureau and was only for occupied units. He used American Community Survey data to match census tracts on these factors:
- Non-Hispanic White share of population
- Non-Hispanic Black share of population
- Non-Hispanic Asian share of population
- Hispanic share of population
- Total population
- Foreign-born share of population
- Other-language speakers (age 5+)
- Renter-occupied housing share
- Poverty rate
- No internet access
- Under-5 population share
(We at Count All Kids are particularly grateful that he took into account the under 5 population.)
He looked at more than 83,000 Census tracts, in all 50 states, which we consider to be a very robust sample. The matching process removed 11,076 tracts. The final matched sample of tracts of 72,262 tracts included 45,699 “treated” tracts with state outreach investments and 26,563 “control” tracts with no state investment.
He found that state investment in outreach, of any level, increased self-response rates by 1.46%.
He found that there were 1.1 million additional household self-responses attributable to state outreach across 45,699 matched tracts with state investments.
Professor Rocco’s work also found particular impact in communities of color, communities with linguistic barriers, and communities with high numbers of renters.
To look at the effects of state outreach investments in tracts with different demographic characteristics, he split tracts into four quartiles for each characteristic he cared about and evaluated the size of the treatment effect in each of these quartiles. and then analyzed how big the program’s effect was in each group. The effect of state outreach investments got steadily stronger in census tracts with more people of color residents. It climbed at every step, from about 0.67% in tracts with the lowest nonwhite population share (Q1) to 2.48% in the tracts with the highest nonwhite population share. This is a statistically significant increase in all tracts, but far greater for communities with high numbers of people of color.
When he broke out the results for communities with linguistic barriers, the quartile of communities with the most people with linguistic barriers saw an increased self-response rate of 3.07% compared to the quartile with the least people with linguistic barriers, where the self-response rate only increased by .74%; again in both cases this is statistically significant.
When he broke out the results for numbers of renters, the lowest three quartiles had increased response rates of 1.22%, 1.06% and 1.32%, but the response rate in the quartile with the most renters was 2.10%.
We know from research by Dr. William O’Hare that young children are missed for different reasons than adults, so an increase in self-response would not necessarily mean an increase in counting young children. However, we believe that this research shows promising results for counting young children because it shows increased response rates in some of the same kinds of communities that also have significant undercounts of young children. Young Hispanic children and young Black children were missed at high rates in 2020. Census Bureau research shows correlations between counties with high undercounts of young children and high numbers of renters. Population Reference Bureau research found a correlation between young child undercounts and linguistic barriers, as well as renters and other factors.
We are also delighted to report that Prof. Rocco intends to continue this line of research. Questions he hopes to examine include whether there is a minimum level of investment needed to create a response, what are the causal mechanisms linking spending to response in specific subpopulations, what are the moderating effects of timing and scale, and how state and local investment interactions shaped outcomes. We have also suggested that he look at the following questions. Did state investment improve self-response rates in communities with high numbers of children in poverty, high numbers of single parent households, and high numbers of households where no adult completed high school, all of which were found by the Census Bureau to also correlate with the undercount of young children? Did the response rates improve in communities with high numbers of multi-unit buildings (there were concerns in 2020 that enumerators could not get in such buildings, so being able to improve self-response rates in these facilities would be helpful.) Did response rates improve in larger counties? We know that large counties have very large undercounts of young children.
We will continue to report on his work as he has new findings. In the meantime, we hope these findings helps you get states and communities to form Complete Count Committees and focus on the communities we know respond to their outreach efforts.
