The blog of the Urban Institute
June 13, 2019

Assessing the Census Citizenship Question’s Potential Effects on Congressional Representation

Last month, documents in a federal court filing revealed that former political consultant to the Republican party and top gerrymandering expert Thomas B. Hofeller played a critical role in the attempt to include a question on citizenship status in the 2020 Census. The contested question would ask all persons living in the United States to indicate whether they are a US citizen.

The US Department of Commerce’s official argument for the citizenship question is that it would improve enforcement of the Voting Rights Act of 1965 and result in fairer elections. Some policymakers have said that even if the citizenship question lowers census response rates among Hispanic and Latinx people (an effect we explore in our interactive feature), the benefits to voting rights outweigh those concerns (PDF).

Assessing the possible voting rights benefits of a citizenship question is difficult because the US Department of Justice does not appear to be using citizenship data to support voting rights. But we can explore the political effects of the census citizenship question related to other potential factors.  

1. Apportionment

There is disagreement over whether congressional apportionment should be based on the number of citizens or total population. Currently, allocation of districts between and within states is based on population, which includes noncitizens. Republicans have argued for decades that population-based apportionment is unfair because, even though noncitizens cannot vote in statewide or federal elections, large noncitizen populations can lead to larger districts and give those districts more political clout.

At the national level, the Constitution mandates that the allocation of districts to states be based on the total population. But some courts have indicated that a state can base its internal redistricting on the citizen population instead of the total population only if high-quality citizenship data are available. If a citizenship question is in the census, that would likely enable states to use citizen counts for their internal redistricting.

2. Gerrymandering

Congressional districts are often designed with partisan motivations, a process called gerrymandering. Gerrymandered districts allow one political party to win more districts than it would if districts were based solely on the proportion of the votes received by that party. The basic technique in gerrymandering—which both Republicans and Democrats have used—is to draw many districts for the favored party to win by a slim margin and a few districts for the opposing party to win by a landslide. Citizen-based reapportionment can be a useful tool (PDF) for concentrating opposition voters into even fewer districts.   

To measure how the citizenship question might affect these two factors, we designed a model of how Texas (which has a large population of Hispanic, Latinx, and immigrant residents and, by some measures, is considered one of the most gerrymandered states in the US) could be affected by citizen-based versus population-based redistricting, with and without gerrymandering factored in. Our model (PDF) uses county voting patterns from the 2016 presidential election and the citizen populations of each county from the 2013–17 American Community Survey.

Here is what we found:

figure 1


Texas had a majority of Republican votes for the 2016 presidential election. Population-based districting with no attempt at gerrymandering (the first yellow bar), is expected to give the state a slightly smaller majority of Republican districts than the vote would indicate.

Citizen-based districting with no attempt at gerrymandering (the second yellow bar) is expected to give Texas about the same share of Republican districts as the vote would indicate. Because noncitizens aren’t allowed to vote in federal or statewide elections, citizen-based apportionment would match the actual distribution of votes in the absence of gerrymandering.

But we also see that when gerrymandering is factored in, it would give a substantial majority of Republican districts well beyond their voting shares. Such a disparity (the difference between the blue and yellow bars) would be many times the magnitude of the gap between citizen- versus population-based apportionment (the difference between the two yellow bars).

Furthermore, switching a gerrymandered state to citizen-based districting would enhance the effectiveness of the gerrymander. In Texas, with gerrymandering factored in, switching from population-based to citizen-based apportionment would result in an almost 6 percent increase in Republican districts (the difference between the two blue bars). This translates to a swing of 2.2 congressional districts from Democrat to Republican out of the 36 congressional seats. 

This chart also shows the 2016 distribution of congressional seats in Texas. The fact that the share of actual Republican districts in 2016 is so much higher than the share of Republican votes indicates that gerrymandering plays a large role in how districts in Texas are currently drawn.

Our accounting of the possible consequences of a census citizenship question suggests that in the absence of gerrymandering, citizen-based apportionment could be defended as fair. But citizen-based apportionment can enhance gerrymandering’s effects.

This political effect is just one of many potential results of the proposed citizenship question—including lower response rates among Hispanic and Latinx residents and, consequently, less federal funding for communities with large populations of these groups. A fair census count and equitable political representation can bring the country one step closer to dismantling existing structural disparities and a more just future.

Photo by Anadolu Agency / Contributor / Getty Images.


As an organization, the Urban Institute does not take positions on issues. Experts are independent and empowered to share their evidence-based views and recommendations shaped by research.