If a citizenship question is included on the 2020 Census, more than 4 million Americans could be missing from the final count, according to a new report. Washington’s population could be undercounted by about 1%, or more than 75,000 people.

The report, which estimates potential inaccuracies in census counts for all 50 states and the District of Columbia, was published by the Urban Institute, a nonpartisan Washington, D.C.-based think tank.

A citizenship question hasn’t been included on the decennial census since 1950. The Trump administration approved adding it back, stating that it intends to use the data collected to better enforce the Voting Rights Act.

Critics aren’t buying that explanation, insisting the decision is purely political. Indeed, the Census Bureau’s own research shows that the inclusion of a citizenship question will scare some households with noncitizen members, suppressing participation. And recently unearthed documents from the computer hard drive of a deceased Republican strategist spell out plainly how a citizenship question could help Republicans and hurt Democrats.

A coalition of states and cities filed suit to block the administration from asking the question. The Supreme Court is currently deliberating the case, and a decision is expected this month.

“The citizenship question is going to have more of an impact on certain subpopulations, like the immigrant community, the Latino or Latinx/Hispanic community, Asian populations, and racial or ethnic populations that have higher levels of refugees and immigrants,” said Rob Santos, vice president and chief methodologist at the Urban Institute.


Even if the question is barred from the 2020 Census, a lot of the damage has been done already, Santos believes.

“Regardless of what the Supreme Court decides, that impact is still going to be there,” he said.

The report offers three potential scenarios for the 2020 Census. The scenario that takes into account the inclusion of a citizenship question shows the most severe undercount.

For Washington, the 2020 Census could undercount both the black and Hispanic populations by nearly 4%, according to the report. The Native American population could be undercounted by about 2% and the Asian population by about 1.5%.

The white population is also at risk of being undercounted, but to a much lesser degree — in the worst-case scenario, by less than .2%.

The biggest potential undercount — nearly 6% — could be for Washington children under age 5.


There are other issues beyond the citizenship question that threaten the accuracy of the  count. The Census Bureau has introduced various operational changes for 2020, but Santos says they’ve been insufficiently tested due to a lack of money.

“Because of underfunding, and therefore the need to do less testing than the Census Bureau could have or should have, there may well be impacts,” he said.

Additionally, there are certain populations that are particularly hard to count. These include people who live in complex households, like multigenerational households, or different, unrelated families pooling resources. Low-income households are harder to count because their members tend to move more frequently, and sometimes miss the opportunity to fill out the census form. Renters are generally harder to count than homeowners. While all these groups have always existed, they now make up a greater share of the population than they did in 2010, so the undercount could be greater, Santos says.

Meanwhile, there are populations that actually could be overcounted. For example, in Washington, people 50 and older are more likely to be overcounted than undercounted, according to the report.

How do some people get counted more than once?

There are a number of scenarios. For example, a household that owns a main residence and a vacation home might return a form from each address. For divorced folks who have joint custody of children, sometimes both parents include the kids on their census forms, and the children can get double-counted. The elderly are also more likely to submit census forms twice.

The Census Bureau tries to capture errors such as duplication, and remove them, but it’s not perfect.


For states and places with a higher percentage of harder-to-count populations — immigrants, people of color, renters — the risk of an undercount is greater. The ramifications are significant.

Census population figures are used to determine the number of congressional seats and Electoral College votes each state receives. An area that’s undercounted could have a diminished political voice in the U.S. Congress.

Census numbers are also determine how billions of dollars in federal funds are distributed.

“It’s a zero-sum game,” Santos said. “There’s X amount of funding that gets pushed out to states, and if some states are getting less than they should, that means some states are getting more than they should.”

Santos says it’s possible that the 2020 Census’s undercount could be worse than the Urban Institute’s report predicts.

“We only quantified things for which there was hard research, where we could say, ‘OK we see this, therefore we think it’s going to have this impact,’ ” he said. “There may be some impacts that we don’t know how to quantify.”

A lot of states could fare even worse than Washington — in particular, California and Texas face potential undercounts of close to 2%. For Washington, D.C., that number could be closer to 3%.

Some states — ones where white people make up a larger share of the population — have much lower risk of an undercount. And there’s one state wheres the most likely scenario is an overcount: Vermont.