Bad Faith and the US Census

Michael Stephens | January 18, 2019

A federal judge has ruled that the Trump administration’s attempt to add a citizenship question to the next (2020) decennial census is illegal. The administration has already begun the process of appealing the ruling.

One way to understand the broader context behind this proposed change is to see it as part of ongoing attempts to influence the outcome of the democratic process (efforts which include gerrymandering, voter registration purges, and so on). In this case, the addition of the citizenship question would lower response rates (that is, lead to an undercount of the population) in areas with higher proportions of immigrants — documented and undocumented. These lower response rates, in turn, could affect the apportionment of congressional seats — that is, reduce the number of seats representing those areas of the country (the undercount would also reduce the level of federal funding directed to such areas).

And as the one-pager below by Senior Scholar Joel Perlmann makes clear, the ostensible justification for this change — to obtain citizenship data in order to enforce the Voting Rights Act — is weak, when weighed against the aforementioned “side effects.”  As Perlmann points out, there are ways to obtain this data that do not undermine the integrity of the full census count. Unfortunately, the latter is most likely the entire point of this endeavor.


A Citizenship Question on the US Census: What’s New? (pdf)

by Joel Perlmann

By now most readers will know that the Trump administration and Commerce Secretary Wilbur Ross in particular have been pressing for a “reinstatement” of a question on citizenship status to the US census that will be conducted in 2020 (Commerce oversees the Census Bureau). There has been a good deal of pushback. Opponents have cogently argued that asking the question today will encourage immigrants to fear that filling out the census form will endanger their residence in the United States. This reaction is easy to understand for the undocumented, but it also applies to immigrants who have arrived through the legal processes for immigration. Many in this latter group will be unsure of how the question will be used against them and decide that the safer course is to ignore the census. Reduced census counts for immigrants, in turn, will reduce both federal aid to local areas and (especially) the number of congressional seats allocated to states with larger immigrant populations. Alone, this reduced count may not mean too much. But like other shady mechanisms for skewing representation or pruning the numbers of eligible voters, every little bit has an impact.

Still, if the citizenship question was asked in the past, why not reinstate it now? The answer turns on historical insights: the characteristics of immigration and of the census itself have changed radically in the meantime.

The US Constitution mandates an enumeration of all individuals (not just all citizens, by the way) once a decade in order to establish each state’s seats in the House of Representatives. From the beginning, Congress saw the advantages of asking for more than just a total count, and over time questions about a variety of characteristics of the population were added. In particular, Congress added a question on an individual’s country of birth in 1850 and it has been asked ever since. Then, in 1890, Congress added another question on the citizenship status of the foreign born, and that question was asked in each census through 1950. But after 1950, the citizenship question was always asked of only a sample rather than the entire population. In the recent past, it has been asked of about 1 percent of the population—over 3 million respondents—each year.

Why was the citizenship question added in 1890? There was much debate at the time about whether new immigrants really meant to stay in this country. It was a reasonable question because there were many “birds of passage”: migrant workers who only intended to take advantage of the faster, safer travel on the large ships of the day to spend a season, or a few seasons, in America and return to the old country with savings. Ascertaining whether an immigrant had taken the trouble to gain citizenship was an important part of the answer. Crucially, there was no particular reason at the time to fear that the immigrants would be afraid to answer—because the number of immigrants coming into the United States was essentially unregulated before the early 1920s. Indeed, for much of the period, many states sent recruiters to Europe to stimulate immigration. Beginning in the early 1920s, drastic quotas restricted immigration and so the citizenship question in 1930, 1940, and 1950 was something of a leftover, a question that pertained mostly to an aging immigrant generation that had arrived by 1920.

So one big change between then and now is that the census would today ask the question in a context of highly restricted immigration and many undocumented arrivals. The other big change concerns the nature of the census itself. Starting in the mid-twentieth century, as sampling expertise and computing power were revolutionized, the Census Bureau began to reduce the number of questions asked of the entire population, and to exploit the use of samples addressed to a random selection of the population. By comparison to any other social science project, these census samples are gigantic—the most important include millions of sample members—and gigantic, too, have been the financial savings from not needing to ask a long list of questions of all American households.

Most people would probably be astonished to learn how little useful information, beyond the actual full count of people, is gained today from the census itself—the once-a-decade full enumeration of the population. The decennial census typically includes questions only on age, sex, race, ethnicity, and exact address. All the rich information about schooling, occupation, income, welfare, homeownership, rental costs, and much else comes only from the sample (which also includes all the questions found on the 100 percent enumeration, not least to serve as a check on sampling quality). Since 2000, the most important of these samples has been the American Community Survey (ACS), a yearly 1 percent sample of the population. The citizenship question is always part of the ACS. Are the sampling data detailed enough to allow us to study citizenship at the local level? That’s exactly why the ACS is so large—upwards of 3 million sample members every year or 30 million over the course of a decade.

Given the massive historical changes since 1890, reinstating the citizenship question on the full, once-a-decade census enumeration would be expensive and redundant because of the current heavy reliance on the ACS, and it would serve to threaten the integrity of the census counts because immigration now occurs in the context of sharp legal limits unknown in 1890.

Senior Scholar Joel Perlmann is director of the Institute’s Immigration, Ethnicity, and Social Structure program and the author of America Classifies the Immigrants: Ellis Island to the 2020 Census (Harvard University Press, 2018).


Leave a Reply