Will the citizenship question break the census? No, says economics prof

Matt Weeks | Jul. 09, 2019

The next U.S. Census will be conducted in 2020, providing the federal government with a host of data about American residents. That information will be used to determine things such as the number of seats each state has in the House of Representative and how federal funds are distributed. 

The University of Georgia’s Ian Schmutte, associate professor of economics at the Terry College of Business and former economist at the U.S. Census Bureau, has published research about the agency and how it can best deliver their data to policymakers. 

What kind of information does the government want from the census?

The decennial census is the primary source of information about the demographic makeup of our country. Section 2 of the 14th Amendment to the Constitution requires that we conduct a census of the population every 10 years. Currently, the decennial census collects information on the number of people living in a household along with their gender, race, ethnicity and age.

Using this information, the Census Bureau can publish detailed statistics on the demographic makeup of all neighborhoods (e.g. census tracts) in the U.S.

Data from the census is used:

  • For the apportionment of congressional seats to the states,
  • By state legislatures to draw congressional and other voting districts
  • For the allocation of around $675 billion in federal funds
  • For enforcement of the Voting Rights Act
  • By city planners to help plan, for example, the locations of schools or emergency services
  • By businesses to help choose new locations and predict market opportunities


The decennial census is also used as a “gold standard” basis for the design and evaluation of other surveys conducted in the public and private sectors.


Do taxpayers have access to the data collected?

The Census Bureau provides copious amounts of information based on the census to the general public. It provides literally millions of tables that describe the demographic makeup of each neighborhood in the U.S. In addition, the Census Bureau publishes the anonymized data of a random sample of individuals that researchers, reporters and others can use to assemble their own customized analysis.


There’s been a lot of talk about whether or not a question about citizenship should be included in the questionnaire.  How would that affect the data and how the government releases it?

Whatever questions are included, the Census Bureau will take all necessary steps to ensure a complete count, and to comply with its dual mission to provide accurate statistical summaries of the population while protecting the privacy and confidentiality of its respondents.


How do you balance the right of privacy on the census with the public good that it represents?

This is ultimately a social question that depends on how we, as a polity, balance the benefits of having accurate data with the costs of privacy loss. To be clear, by law the Census Bureau is not allowed to share the information it collects with other parts of the government that might use them for, say, law enforcement or other administrative purposes. When people respond to the census, they should feel very confident that their private responses will not be inappropriately shared with other parts of the government, nor leaked due to a security breach.

That said, my research with John Abowd deals with privacy loss that arises from publishing summaries of individual responses to the Census. Those publications have the potential, when combined with other information, to reveal small amounts of information about individuals. The Census Bureau has long been aware of this issue and uses tools of statistical disclosure limitation to ensure that privacy losses are as negligible as possible, given the need for accurate data.

In our recently published papers, Abowd and I argue that publishing better statistics – which enable better decisions – requires greater risk of privacy loss. To understand whether we should publish more accurate statistics depends on the cost of accuracy, in terms of increased risk of privacy loss and our willingness as a society to pay that price. Both are hard questions to answer, and depend on the value of better governmental and business decisions, and the harm from privacy losses.