The melancholy of having to count souls
Where they grow fewer and fewer every year
Is extreme where they shrink to none at all.
It must be I want life to go on living.
Robert Frost, “The Census-Taker” (1923)
It was built as a quartermaster depot during the Civil War. Later, it became a shirt factory. During the Korean War, parachutes and refrigerated trucks were manufactured there. But for the last sixty years, the massive complex in Jeffersonville, Indiana, close to the Kentucky state line, has been devoted to counting everyone in the US. Last week––in the midst of the Covid-19 pandemic––employees of the Census Bureau’s National Processing Center were advised not to enter the building “until further notice” while it receives a “deep cleaning.”
The temporary closure of the massive Jeffersonville facility serves as a reminder that the knowledge of populations depends on processes that are situated, local, and material. This history is made invisible when the information we provide on forms is processed, tabulated, and compiled into the statistics that structure congressional apportionment, legislative redistricting, and the allocation of $1.5 trillion in federal grants.
Few of us learn about the census in high school or college. You won’t find it on many syllabi in the field of political science. A sizable number of people I know can barely remember filling out their 2010 form.
But crises have a way of inverting infrastructure. Strained to the breaking point, that which is taken for granted becomes painfully visible. Questions multiply. Doubts surface. And just when simplicity would be useful, things become complicated.
The Covid-19 pandemic is not the first rude surprise to greet the 2020 Census. The last year has been a perfect storm. The Trump administration’s attempt to place a citizenship question on the census form — halted by the Supreme Court last summer — has created a climate of fear in immigrant communities. Several successive federal budgets left the Census Bureau fiscally constrained. Critical field tests were cancelled, area census offices downsized, and questionnaire assistance centers eliminated. A tight labor market and more stringent background checks weakened census hiring efforts. Projected undercounts appeared dismal.
Pressures like these have added to the high fiscal and political stakes of the census for state and local governments. In doing so, they have generated a new intergovernmental infrastructure for census counts. State and local governments now collectively invest billions of dollars to mobilize participation in the census through advertising and awareness campaigns executed by nonprofit organizations. At one end of the spectrum, California has begun to build a parallel census infrastructure — creating a state-level office of the census, conducting extensive statewide education and mobilization operations, and organizing its own independent “neighborhoods count” to survey housing units in historically undercounted communities. Not all states, of course, have taken this approach. At the other end of the spectrum, Texas has made no dedicated resource investments in the production of the 2020 Census, leaving local units of government to take on this work themselves. In between these extremes, states have relied on a combination of modest investments, philanthropic partnerships, and nonprofit networks to tailor census outreach efforts to historically undercounted populations.
The intergovernmentalization of census-taking has been the subject of my research for the last six months or so. Before the pandemic hit, I spent dozens of hours on the phone with state and local officials and nonprofit leaders talking about their plans to ensure that people in historically undercounted communities fill out their forms.
Census-taking was not necessarily an obvious subject for someone who studies federalism. Indeed, not only does the Constitution make census-taking an exclusively congressional responsibility, the rise the federal statistical system in the early twentieth century vested crucial intellectual and material capacities within the Census Bureau itself. Yet the intergovernmental dimension of census-taking is a variable, not a constant. Prior to 1902, the U.S. had no permanent administrative infrastructure for the census — relying instead on congressional direction and ad hoc organizational arrangements. At the same time, a great amount of census-taking activity occurred at the subnational level; between 1810 and 1945, 27 states took 160 separate censuses for the purposes of apportionment, taxation, and economic planning. In the absence of robust capacity at the national level, local knowledge and political action also helped to reshape national census-taking practices. The ‘census tract’ itself was conceived of by Presybterian minister Walter Laidlaw and first deployed in the New York State Census of 1915.
Following the development of the federal statistical system, state and local governments took on a more residual role in the decennial census. Nevertheless, following the rise of the federal grant-in-aid system and reapportionment revolution of the 1960s, intergovernmental action on the census has increased. On the one hand, as the stakes of a statistical undercount for federal funding increased, so too did state and local litigation over the Census Bureau’s methodology. Following every census count since 1980, cities and states have challenged census population estimates in federal courts. At the same time, state and local officials have also participated — in one form or another — as partners to the Census Bureau in the production of the census. The creation of the Federal-State Cooperative on Population Estimates in 1967 eventually gave state-level demographers and statisticians a role in reviewing and adjusting census counts. The establishment of State Data Centers in the late 1970s allowed state and local officials to help shape (and repackage) the data products generated by the Bureau. Since 2000, the Census Bureau has formally invited all state and local governments to establish “Complete Count Committees,” which coordinate Get Out the Count (GOTC) efforts. This can involve targeting census messaging to local hard-to-count communities, leveraging the voice of state and local government agencies in census promotion, and convening and financing community groups to execute their own census outreach efforts.
In the weeks leading up to the pandemic, I had a pretty steady routine. At around 9, the calls began. I listened to my interviewees’ stories about impromptu census potluck dinners in Denver, training Philadelphia’s barbers and hair-braiders to talk about the census, and — virtually everywhere — begging for more resources, support, and agenda space.
Quarantine has put a stop to all that, and to a great degree it has put a stop to the types of census mobilization work I’ve been tracking: festivals, block parties, community meetings, door-to-door canvasses, “census Sundays” at mass have all ceased. The Census Bureau has delayed its own field operations. One of my interviewees —a manager in a medium-sized western city — described the last days of his work as a surreal scramble to migrate events online. “I told my wife it felt like Shaun of the Dead,” he said.
I’m asked somewhat regularly what effect the pandemic is having on the count. In one way, the effect is already obvious. Covid-19 has pushed back the Bureau’s operational schedule and virtually every in-person activity undertaken by state and local governments. On the other hand, it’s difficult to nail down the specific effects of the pandemic. Thus far, self-response rates are behind where they were in 2010. Yet if there has been any relationship between cases of Covid-19 and lower rates of self-response to date, the state- and county-level tracking data make it difficult to recover.
If anything, I suspect that strained followup operations and cancelled public-awareness events might have the largest impact in communities that are already historically undercounted. After all, there is far greater variation in census participation at the tract level than at higher layers of the census spine. Here are the tract-level self-response rates in Milwaukee March 30, 2020.
For comparison, here were Milwaukee’s final self-response rates for 2010.
While I haven’t been able to merge the tract-level data on Covid-19 cases, a look at Milwaukee County’s dashboard illustrates that there is at least some visible correspondence between the tracts with the lowest self-response rates and those with the largest numbers of cases.
If you know anything about Milwaukee, you’ll recognize that the neighborhoods most likely to be adversely affected by Covid-19 are also at a higher risk of being undercounted. They’re also the places where the fiscal effects of an undercount — just like the economic effects of a pandemic— are likely to hit the hardest.
It’s too soon to test the effect of the virus on the census. But the effects of enduring racial and economic inequality on every number that matters are already manifestly clear.