For decades, little kids have given the U.S. Census Bureau the slip.

Although the decennial census intends to count everyone in the U.S., demographers suspect large numbers of young children don’t get counted, and the omission comes with a cost.

The census guides the distribution of $1.5 trillion in federal spending divided among states and localities based on population. An undercount in a given area means that place might not get its fair share.

At stake are funds for the Supplemental Nutrition Assistance Program, which helps needy families purchase healthy food; Head Start, which provides early-childhood education and other services to low-income children; the Children’s Health Insurance Program, which offers low-cost coverage to children in families that don’t qualify for Medicaid; as well as money for other federal programs.

“We are finding the most disadvantaged neighborhoods are most at risk of being undercounted,” said Mark Mather, a demographer at the Population Reference Bureau, a Washington, D.C., research center. “What’s important is those are the communities that need the help the most.”

“‘The most disadvantaged neighborhoods are most at risk of being undercounted. What’s important is those are the communities that need the help the most.’”

— Mark Mather, demographer, Population Reference Bureau

It isn’t clear what causes children to be undercounted, and it’s likely that there are several explanations. Some households might be missed entirely. Language barriers could cause inadvertent omissions. Or families might intentionally leave out young children because they don’t believe kids are meant to be counted or because they don’t trust the government.

The group most likely to be missed is children ages 0 to 4, according to demographers who have detected the discrepancy by comparing census tallies with population estimates based on birth and death records.

The net undercount in 2010, they say, was 4.6%, or around one million young children, and preliminary figures hint that an undercount in 2020 could be larger.

According to William O’Hare, author of the book “The Undercount of Young Children in the U.S. Decennial Census,” young children have been undercounted habitually, but the gap has increased.

“From 1950 to 1980, the net undercount was about the same as most other demographic groups and was improving,” Dr. O’Hare said. “Since 1980, the rest of the population has continued to improve, but the undercount of children has gotten worse.”

In 1980, the net undercount of young children was 1.4%, according to his calculations, more than 3 percentage points lower than in 2010.

Potential over- and undercounts are revealed when the Census Bureau and others check the accuracy of the survey using two primary methods: a post-enumeration survey and a demographic analysis.

A post-enumeration survey, conducted immediately after the official count, asks a sample of the population the same questions as the census. The 2020 sample included approximately 180,000 housing units.

The responses are matched case-by-case to the census, and differences between the two are used to evaluate how many people were likely missed or double-counted the first time around. In addition to testing the accuracy of the census, the information is used to improve future counts.

When Dr. O’Hare compared the 2010 census and post-enumeration survey, he found only a 0.7% net undercount of young children. But because people who are missed in the census might also be omitted in the follow-up survey, Dr. O’Hare said a better measure is the demographic analysis, which in 2010 revealed the net undercount of 4.6%.

Demographic analysis calculates population sizes using birth and death certificates, estimates of net international migration and, for people 75 and older, Medicare enrollment.

“The demographic analysis estimates are very good, and they are particularly good for children,” Dr. O’Hare said. “Four-year-olds are born here, almost none die, and hardly any immigrated.”

The most recent demographic analysis was released in December.

Census counts of young children, and other detailed information, won’t be released until next year, but to get a sense of how well children, in general, were represented in 2020, Dr. O’Hare compared the number of children ages 0 to 17, which was released in August, with corresponding figures from the demographic analysis.

SHARE YOUR THOUGHTS

Do you have young children in your home? If so, did you report them on the census? Why or why not? Join the conversation below.

The demographic analysis identified 74.6 million. (Dr. O’Hare used the middle of three estimates; the low and high numbers were 74.2 million and 74.8 million.) The census recorded 73.1 million, suggesting an undercount of 1.5 million.

But the demographic analysis also has issues that could affect its accuracy.

In 2010, approximately 200,000 young children who were born in the U.S. were living in Mexico at the time of the census and were not accounted for in the demographic analysis, which would reduce the size of the undercount. In 2020, the bureau took steps to better measure this population, and it also worked with pediatricians and child advocates to reinforce the idea that everyone in a household, regardless of age or relationship to the householder, should be counted.

Because when the Census Bureau can’t find children in this high-stakes game of hide-and-seek, the kids are always the losers.

Write to Jo Craven McGinty at Jo.McGinty@wsj.com