Skip to main content
ABC News
The Citizenship Question Could Cost California And Texas A Seat In Congress

In a matter of days, the Supreme Court may dramatically change the census. The court is slated to rule on whether the Trump administration can add a question about citizenship to the 2020 form. When the case was argued back in April, many court-watchers predicted that the court’s five conservative justices were ready to side with the administration. The proposal sounds innocuous enough, but social scientists and civil rights advocates worry it will deter vulnerable populations — particularly undocumented people, other immigrants and their families — from answering the census. If that happens, many people from these groups will be at risk of not being counted and huge swaths of American life will be affected. The results of the count determine everything from where grocery stores are placed to how congressional representatives are distributed.

There are few things we care more about around here than political apportionment (although, if we’re being honest, we care an awful lot about groceries, too). So we went in search of researchers who had estimated the potential effect of the citizenship question. We found several, none of whom agreed on just how big an impact this would have. But they were all on the same page about one thing — if the Supreme Court rules that the new question can be included, it could alter our political future.

Every 10 years, the updated census numbers are used to determine how many U.S. House members each state will get. So figuring out who might be missed and where can tell us a lot about who stands to gain political representation and who stands to lose. Forecasting that amounts to sophisticated guesswork, since the question hasn’t yet been field-tested by the Census Bureau — a decision that many experts regard as a scientific cardinal sin. But that hasn’t stopped researchers from trying to fill the void.

I talked to three political scientists who all took a stab at quantifying how much each state’s population would change if some of the people who seem most likely to be affected by the question — immigrants, noncitizens and Latinos — simply went uncounted. All three used different assumptions about how people will feel when they’re confronted with the questionnaire — a kind of statistical choose-your-own-adventure game. You can see five estimates calculated by those researchers in the chart below:

These estimates make different assumptions about who will be missed by the census, which led to fairly substantial differences in how evenly the undercount was distributed and which states were most affected. Estimates for population loss in California alone ranged from almost 700,000 to over 1.8 million. Notably, the Census Bureau’s own estimate (Scenario 1) was by far the most conservative. Let’s take a closer look:

  • Scenarios 1 through 3 come from George Washington University’s Christopher Warshaw,1 a political scientist who ran two analyses based at least in part on the Census Bureau’s estimate of the question’s effect, and a third based on his own survey experiment.2
  • Scenario 4 comes from Bryce Dietrich, a research fellow at the Harvard Kennedy School’s Shorenstein Center on Media, Politics and Public Policy. Dietrich worked with the Washington Post to estimate the impact that an undercount of about 6 million Hispanics would have on state populations.
  • Scenario 5 comes from political scientist Eric McGhee of the nonpartisan Public Policy Institute of California, who estimated what would happen if the census was poorly conducted — a distinct possibility, given that the Census Bureau is perpetually strapped for cash — and 10 percent of households that include undocumented immigrants were missed.

The question for many political observers, of course, is what all of this will mean for what happens after the census, when the states are assigned seats in the House of Representatives based on their new population count. Some clear winners and losers emerge here.

California would lose seats in all five undercount scenarios

Estimates of how an undercount would affect the apportionment of U.S. Representatives after the 2020 census, for states where any impact was estimated in five scenarios from researchers

Difference in number of seats due to undercount
undercount scenario AL AZ CA FL MN MT OH TX
1. 5.8% of noncitizens undercounted 0 0 -1 0 0 +1 0 0
2. 5.8% of noncitizens and Hispanics undercounted +1 0 -1 0 0 +1 0 -1
3. 5.9% of Hispanics and 11.3% of foreign-born non-Latinos undercounted +1 0 -1 -1 +1 +1 0 -1
4. Poorly conducted census and 10% of households with undocumented immigrants undercounted 0 -1 -1 0 +1 +1 +1 -1
5. 6 million Hispanics undercounted +1 -1 -2 0 +1 +1 +1 -1

1. Based on Census Bureau estimates from Aug. 2018.

2. Partially based on Census Bureau estimates from Aug. 2018.

3. Based on a survey experiment that tested how much less likely respondents were to say they would return the census form if the citizenship question was included.

4. Partially based on the undercount in the 1990 census.

5. Based on a survey experiment that tested how much less likely respondents were to respond to individual questions on the form if the citizenship question was included.

2020 population projections were derived by the researchers and vary from scenario to scenario.

Sources: Christopher Warshaw, Bryce Dietrich and The Washington Post, Public policy institute of California

California loses at least one seat in all five scenarios, while Montana stands to gain one. Ohio, Minnesota and Alabama are all potential winners, depending on the scenario, while Texas, Arizona and Florida come out on the losing side in at least one estimate. “This is a fairly significant impact, especially when you consider that some of these states would stand to gain seats if not for this question,” Warshaw said.

We can’t really tell you which scenario is most likely to play out (assuming the court opens the door to any of them). It all hinges on which groups are ultimately missed in the final count and by how much. The actual effect of the citizenship question could be much larger than any of these estimates, if other groups refuse to respond as well. Or it could be smaller, if the Census Bureau is able to track down some of the people who don’t respond to the first questionnaire they send out. Most of these estimates are based on the idea that every person who doesn’t initially respond to the census, or who skips a question or two, won’t be counted. In reality, at least some of those people will almost certainly be included in the count because armies of census workers will fan out across the country next year to find people who didn’t respond to the initial request for information.

As the lawsuits over the citizenship question unfolded, Census Bureau officials argued that they will be able to respond to some of the fears swirling around the question through community outreach and follow-up. In an email, a Census Bureau spokesperson told me that in response to concerns about an undercount, the agency is “developing a robust communications campaign and working with communities across the country to communicate that responding to the census is safe, easy, and important.” The agency is also about to begin a test of the question that it says will help determine how many census workers are needed to follow up with people who don’t respond at first. Rob Santos, a researcher at the nonpartisan Urban Institute and the co-author of a recent report on miscounts and the census, said that these efforts could make a difference, but added that the intense publicity surrounding the citizenship question has already created an “atmosphere of fear and mistrust” in immigrant, refugee and Latino communities.

Nancy Mathiowetz, an expert on survey research, was also skeptical that the census’s follow-up efforts could fully mitigate the effect of the question. She pointed out that Latino citizens and undocumented people are already likely to be undercounted, and it’s not clear if the Census Bureau has any new tricks up its sleeve for reaching them. Some of the impact may also depend on the type of undercount that occurs. Failing to fill out part of the questionnaire, she said, might be less of a problem than leaving household members off an otherwise complete form, since the Census Bureau has said that it will try to fill in certain types of missing information using statistical techniques and administrative records, but it has no way of knowing if family members have simply been omitted. Several experts noted that conjecturing about a respondent’s race and ethnicity could create other kinds of data accuracy problems that could resurface when congressional and state legislative districts are drawn, but the issue for state-level apportionments is simply whether everyone is included in the total count — not whether all the information on each person’s form is correct.

At the end of the day, all of this sparring is a poor substitute for the rigorous, lengthy testing process that a new census question would usually undergo. If the citizenship question ends up on the questionnaire, another test of sorts will unfold in real time, with the entire American population as its subjects. And it will have concrete consequences for who gets political representation. To many social scientists and civil rights advocates, this seems like a needlessly dangerous experiment, and one that runs counter to the cautious, scientific spirit of the federal government’s largest statistical agency. But the Supreme Court may allow it to happen anyway — and if it does, we will soon learn more about just how resilient one of our nation’s oldest civic traditions really is.


  1. In one of the lower court trials, Warshaw submitted an expert report on behalf of the challengers to the administration’s plan.

  2. Warshaw used a forecasting model to estimate what the U.S. population will be in 2020. The other researchers used slightly different methods to estimate the 2020 population, with slightly different results.

Amelia Thomson-DeVeaux is a senior editor and senior reporter for FiveThirtyEight.