Photo of 2010 Census Questionnaire

President Trump’s abrupt decision to drop the citizenship question from the 2020 census is a relief for those who rely on census data for everything from research to allocating federal funds to determining electoral representation. But some experts are warning the numbers will still be off.

If the citizenship question had been included, the Urban Institute estimated that Virginia’s results would’ve missed as many as 94,000 people. But even without it, researchers estimate an under count of 67,500 people due to new and untested data collection methods, a lack of congressional funding, and shifting demographics.

What does that mean? The census count, conducted every ten years, is used to determine states’ representation in everything from the electoral college to the amount of funding they receive from over 300 federal programs. If the tally is wrong, it matters.

“The folks who are under counted won’t get as much as they deserve nor the political representation they deserve, and the folks who are over counted are going to get more than they deserve because it’s a zero-sum game,” said Rob Santos, vice president and chief methodologist at the Urban Institute, in an interview.

Counting inaccuracies have severe consequences for states like Virginia, which received over $17 billion in 2016 for 55 federal programs. That amount was guided by data from the 2010 census.

“The final whammy is that those numbers, once they’re set and official — regardless of their accuracy — are set for the next 10 years and drive things like infrastructure, where stores are placed, schools are developed, and fire stations and so forth are placed,” Santos said.

“That creates a more and inequitable society, and we all lose when there are baked-in structural inequities like that.”

In terms of demographic data, the federal government takes the census as a “gospel of truth,” Santos said — census data serves as a baseline demographic metric for the next decade.

The Census Bureau does do updates between census years, using migration estimates and birth and death records. But if the baseline data is off, it throws the calculations off until the next census is conducted.

“We’ll get closer the quote-unquote real or accurate number as we get further out, but starting from the wrong place definitely does no one any favors,” said Doug Strane, a research project manager at PolicyLab at Children’s Hospital of Philadelphia.

In an interview, Strane said programs like Medicaid, the Supplemental Nutrition Assistance Program, and Medicare Part B would be most affected by a miscount. Virginia collects over $7 billion per year from those three programs alone.

Medicaid in particular relies on census data to determine the amount of funding states get to run the government health insurance plan for low-income groups. If the population estimate is low, the state gets less money than it needs to implement the program effectively.

In states where demand for Medicaid exceeds the dollars coming in to pay for that demand, it could put them in a “tight position where they could look at having to limit or freezing enrollment,” or “limiting eligibility,” Strane said.

That could halt Virginia’s progress on expanding Medicaid — almost 300,000 people enrolled this year after a Democrat-led expansion plan took effect this year.

Children’s health programs would also suffer from a miscount, like the Special Supplemental Nutrition Program for Women, Infants, and Children, also known as WIC. This program divvies out federal money to states based on their populations to cover the cost of food, health and nutrition education for low-income pregnant women, new moms, and children up to age 5 who are at risk of nutritional deficiencies.

The same is true for the low-income heating assistance program, which provides economic support for families during cold-months. It “basically keeps children safe during the winter,” Strane said.

Strane sees the effects of a miscount in his work, like faulty data from the 2010 census that under counted the number of people who rent their homes by 1.1%; over counted white, non-Hispanic residents by 0.8% and under counted 2.1% of the black population, according to the U.S. Census Bureau.

The 2020 census, Strane said, is likely to be even less accurate, even without the citizenship question, due to a lack of funding, new and untested methods of data collection and changing demographics.

Normally, the Census Bureau would conduct three dress rehearsals before a new census goes into the field. This year, due to funding cuts, two of three dry-runs were canceled for the 2020 census. That’s especially problematic because the 2020 census is implementing a host of new data collection methods, like the ability to respond online.

And though Trump’s citizenship question is nixed, the damage might already have been done.

“There is going to be continuing mistrust and other issues that will stifle or dampen participation rates,” Santos said.

The best hope for a better count, both Santos and Strane agreeed, is a grassroots efforts to encourage participation, especially in low-income communities, rural and harder-to-reach areas, and communities with a disproportionate undocumented population. By law, the Census Bureau is not allowed to share data with immigration enforcement officers.

“Trying to demystify the census and make sure that you know, we’re counteracting any misperceptions or misinformation about it is going to be really important to making sure that we get an accurate count,” Strane said.