What I Saw as a 2020 Census Worker

Counting every person in the country is already a massive challenge, but the census was no match for this year’s chaos.

Art of Donald Trump superimposed on a census form.
Shutterstock / The Atlantic

In the coming months, after years of ground-laying, controversy, and anticipation, the United States will finally complete an imperfect civic process that, though heavily compromised by geography, logistics, and partisanship, will affect the life of every single American for years to come. Also, the country will inaugurate a new president.

The 2020 census, which ended its data-collecting operations last month, seems destined to become one of the most telling artifacts of this very strange year. As a census taker, I got a front-porch view of what happens when a 230-year-old national rite runs headlong into a country all but at war with itself and its institutions. Throughout six weeks of door-knocking in dense New York neighborhoods and Georgia backwoods, I got a good look at 2020 America—not only its demography, but also its deep mistrust of government and diminishing faith in the common good.

It’s true that each decennial count encounters its own set of challenges. In a letter to fellow Founding Father Gouverneur Morris, President George Washington sighed that he believed the first census, conducted by U.S. Marshals on horseback in 1790, had produced a major undercount that posed a potential threat to national security. Data from the 1880 census took nearly eight years to compile. Congress threw out the results of the 1920 census entirely.

But the obstacles facing the 2020 census were historic. Even before the counting started, the Trump administration demanded that the census include a citizenship question, a controversial move that set off a yearlong legal battle in which even the Census Bureau warned that response rates would likely decline. And although the Supreme Court stymied the White House’s efforts to add the citizenship question, other barriers—including unprecedented politicization, massive natural disasters, widespread civil unrest, funding shortfalls, high employee turnover, ever-shifting deadlines, and nearly ceaseless litigation—undermined the entire census project. A pandemic hit too. But even a messy census reveals a lot about America in ways that reach far beyond the data.

In February, weeks before COVID-19 and social-distancing mandates interrupted life as we know it, I applied to become a census taker. Every writer jokes that they’ll do basically anything to get out of having to write, but I’d moved to Yonkers, New York, a few months earlier, and also thought the work would be a good way to get to know my mysterious new hometown.

In an ideal census count, all households would submit their own information, which is by far the most accurate way to account for a community’s true demographic makeup. Failing that, the work falls to a less reliable combination of federal administrative data, calculated guesses, and Nonresponse Followup operations, the door-knocking census takers best known to some Americans by an unfortunate reference in The Silence of the Lambs.

My interviewer explained that census training and orientation, and the work itself, were meant to begin in May, after the Census Bureau’s push to get direct responses from households wrapped up at the end of April. We all know what happened next. Following months of pandemic-related delays, not until early August did a census administrator swear me in as a Commerce Department employee in a room of the Yonkers Public Library’s main branch that, in better times, would have doubled as a Zumba studio.

That lag between early May, when door-knocking was supposed to start, and August, when it did, will matter. April 1 is not just Census Day in the way that April 12 is National Grilled Cheese Day. Accurately completing a census case means knowing who lived at an address on April 1, 2020, whether that information is taken from a resident or, oftentimes, a neighbor. The further you stray from the reference day, the less accurate the data become, particularly in a time of heavier population displacement.

As my fellow enumerators and I descended upon our communities in late summer, we were presented with the challenge of sorting out the composition of neighborhoods that had rapidly changed over the course of more than a few tumultuous months. We would also be serving as representatives of a government that had done little to stoke goodwill or faith in the census process.

One tragedy of the 2020 census is that it was supposed to be better. What was meant to separate this decade’s grand enumeration from others was the fury with which community leaders (Cardi B included) and outreach groups intended to push and organize for an accurate tally, especially in hard-to-count areas. The Census Bureau’s own data show it has often failed to fully count communities of color, non–English speakers, lower-income families, immigrants—some of the same demographic groups that have disproportionately suffered during the pandemic.

“The COVID crisis really became its worst at the exact moment when all of those efforts were ready to go,” Steven Romalewski, a researcher at the Center for Urban Research at the City University of New York Graduate Center who closely tracks census response rates, told me. “All of these groups were, in mid-March, ready to hit the streets and go out to all these big, large, in-person gatherings, and the Census Bureau was planning a massive public-relations and advertising campaign. Then, all of a sudden, none of that could be implemented. It was completely upended. And so we’ll never know if those efforts would have dramatically boosted participation in the census.”

The true stakes of a census count, which include congressional representation and more than $1 trillion of annually allocated taxpayer money, are nearly impossible to grasp. According to a report by the George Washington Institute of Public Policy, each person not counted by the 2010 census cost the state they were living in $1,091 in 2015, on average. (For a household of five, that adds up to $54,550 over a decade.) In an interview during this year’s count, a state-census-office manager likened one nonresponse to “walking past a $5 bill every day for a decade.”

During the intensive training process for enumerators, instructors trained us to highlight the financial benefits of census participation. Generally, it was easier to put them in terms like federal money for local roads and school lunches, job-training programs and Medicaid. But getting that message across isn’t easy, especially when political tensions are running high, the government is unpopular across the spectrum, and the census has a partisan sheen.

Resistance to census participation transcends age, race, geography, and party affiliation. Census self-response rates have declined since the 1970 and 1980 censuses (78 and 75 percent, respectively), ranging at or just below 67 percent since the 1990 survey.

In 2018, nearly a third of potential respondents said they were, at best, “somewhat likely” to fill out the census, according to a comprehensive study the Census Bureau completed that year. But that document couldn’t have anticipated that 2020 door-knocking operations would take place in the midst of a pandemic, an extended period of civil unrest and protest, and the closing weeks of a brutal election season.

These obstacles felt especially acute in Yonkers, an immediate neighbor to the Bronx that is one of the most economically and ethnically diverse midsize cities in the country. Prior to the launch of door-to-door visits nationwide in August, the city’s census response rate sat at 55 percent, well below the national average.

For the first few weeks of my assignment, I knocked on doors within a few square miles of my apartment in northwest Yonkers. This section of the city is an expanse of winding, suburban-looking streets, full of Cape Cod houses and multi-family homes, vinyl siding and above ground pools, and well-kept front yards with Virgin Mary figurines. Among its jumble of constituencies, it boasted a surprising number of cars with bumper stickers that offered simultaneous approval for the Grateful Dead and Donald Trump.

In those first few hundred cases, I aspired to crack the census code. I found that responses tended to be better when the weather was good or on weekend afternoons. In the days following a much-delayed pandemic haircut, my close rate jumped significantly. And, despite being a lifelong Astros fan, I bought a Yankees face mask in the hopes that it might boost my trustworthiness among residents. Nevertheless, appearing on a stray doorstep, masked and sweaty in khaki shorts and New Balance sneakers, with a clipboard, a lanyard, a badge, and a list of probing questions, made me feel like the kind of clumsy suburban-dad narc that I myself would probably choose to ignore if given the chance.

What’s funniest about trying and failing to persuade someone to give you 10 minutes of their time for the census is that an enumerator has to document the reason given for a refusal. One rationale is that it gives the next person who attempts to bug a stubborn case a sense of what might be coming. In our data-capture app, many of the prefilled explanations we must enter for why we failed to gather data are unusually blunt: The respondent “does not want to be bothered”; thinks the “survey is a waste of taxpayer money”; has “privacy,” “COVID,” or “anti-government” concerns; or is simply “too busy.”

Attempt enough door knocks and you’ll inevitably run into each of these responses, plus a few others of varying tenor. One guy asked me to come back tomorrow every single time I saw him. A woman standing on a “Live, Laugh, Love” doormat told me to “grow up and get a real job.” One Sunday-morning interview was cut short when the interviewee’s ex-partner appeared early for a custody exchange, which set off an argument. As census efforts intensified in Yonkers, more residences would be marked as a “dangerous address” after a hostile exchange or an unsafe encounter, appearing on the digital map as an ominous red triangle. Over the weeks, the map of many neighborhoods would go from a chance blemish to full-blown acne.

But even if you manage to get someone to answer the census’s questions and assure them of the confidentiality of their answers, you’re still asking for information that seems strange for the government to want to know. You have to ask each person for their phone number; every household member’s date of birth; whether they identify as male or female; whether their children are biological or adopted; whether they rent their home or own it, either with a mortgage or outright. Enumerators are exhaustively trained to expect resistance and quickly display understanding toward any misapprehensions. In census-training parlance, this is known as the “A+ Model,” in which we acknowledge their concerns, answer those concerns by explaining why the information benefits the community, and ask them for their help in completing the census. The A+ Model remains an ambitious formula, especially when most situations better qualify as a simple pass-fail.

Despite our time-tested arsenal of persuasion tactics, even if residents were clearly home, they often didn’t come to the door. I’d be willing to bet my entire census pay that at least half of the 59 percent of census nonresponders who told the Pew Research Center in July that they would be at least “somewhat willing” to open the door to a census worker were either just being nice or outright lying. If there was one boon to the 2020 census process, it was the pandemic-fueled boom in e-commerce sales, which meant that more residents appeared expecting to sign for a package, only to find an artificially cheery enumerator on their landing instead.

On weekly conference calls, my fellow enumerators would air their frustrations about the people visible through front windows who simply wouldn’t come to the door, about being sent repeatedly to addresses that didn’t exist, about the people who were tired of census takers coming to their door. And, above all, about the proxies.

After a certain number of attempts on a case, enumerators are instructed to find a proxy—a neighbor, a mail carrier, a building manager, anyone vaguely credible—to speak on the composition of the residence in question. And in many cases, with enough luck, patience, or cajoling, somebody helps fill in the most basic blank of the census: how many people live at an address. It’s fair to say that this arrangement isn’t the sturdiest blueprint for democratic representation.

According to Terri Ann Lowenthal, a former staff director for the House census oversight subcommittee, 22 percent of cases completed by census takers in 2010 were done so using data taken from proxies. And of those cases, roughly a quarter were deemed useless by the Census Bureau. As a result, millions of people get missed while others get counted twice. These inaccuracies tend to be more frequent in urban centers and tribal areas, but also, as I eventually learned, in rural sections of the country.

Throughout the door-knocking operation, the 500,000 or so enumerators would receive various official updates on our census-branded iPhone 8s, informing us of overtime-pay availability, reinforcing protocols about masks and political neutrality, and letting us know how the ongoing legal action about the census time frame would affect our goals. Over Labor Day weekend, I received a text from our regional office asking for volunteers to travel to southern states, where response rates had been extremely low, especially in rural areas. To take on the assignment, an enumerator would need to commit to following a 13-page document of protocols and have a credit or debit card with enough leeway to hold charges for a rental car, gas, and hotel incidentals until reimbursement.

As many census processes go, my journey to southern Georgia was a bit haphazard. On Labor Day, I received a call from a census director checking to make sure that I qualified for travel. Two days later, without any warning, I received my assignment in the form of hotel and flight confirmations. Thirty-six hours later, on the morning of September 11, I flew out of LaGuardia to Atlanta to join the more than 2,000 other out-of-state enumerators spread out across Georgia.

As it turns out, the mass mobilization of out-of-state enumerators is not just uncommon, but generally seen as a violation of the spirit of the census. “One of the foundational concepts of a successful door-knocking operation is that census takers will be knowledgeable about the community in which they’re working,” Lowenthal explained. “This is both so they can do a good job, because they’ll have to understand local culture and hopefully the language, but also so that the people who have to open their doors and talk to them have some confidence in them.”

Upon arrival, I learned that I had been assigned to the “Save Macon Operation,” which centered around the various farmlands, exurbs, trailer parks, small towns, and off-grid residences to the west of Macon and south of Atlanta. We were led by a kind-but-overwhelmed field manager who had suddenly inherited a squad of 900 foreign enumerators after the local leadership around him had been fired for poor performance. Markus, we’ll call him, often spelled out his name using the NATO phonetic alphabet.

During an introductory conference call, as we went through a litany of payroll and logistical guidances, one visiting census taker interrupted with a question that launched a hundred silences: “What do census takers from the cities need to know about working in rural areas?” A lot.

Dogs and a lack of leash laws would complicate our journey. The GPS would fail often in this part of the state, and the roads themselves would be hard to drive. “You’re gonna take dirt roads to dirt roads to another dirt road,” Markus warned, “and then you’re gonna find half a mile or so of washed-out driveway to get to the house, and there’s probably not gonna be nobody there.”

Bathroom availability, especially with the pandemic, would be limited. We were to expect that proxies would be impossible to find. “These people live way out in the country,” Markus explained. “They want to be left alone.” Then, there were the snakes, fire ants, and chiggers, a revelation that caused one member of the Connecticut delegation to shout, “What in the heck is a chigger?” As Markus sagely explained, “If you get bit by one, you’ll know.”

In a debrief call later in the week, a census taker on loan from Florida brought up an issue he had with a local case where he had arrived at a house to find a sign on the door explaining that because gun ammunition is expensive, the homeowner doesn’t fire warning shots. The enumerator had opted not to knock. “You’re gonna see a lot of that,” Markus said. “These backwoods people, they love their guns. I love my gun. If you feel uncomfortable doing the case, don’t do the case.”

At least three of my fellow enumerators reported having guns pulled on them while entering a property to conduct an interview, one field supervisor told me on condition of anonymity. That two of them successfully managed to complete the cases in question speaks to another crucial feature of the census: In spite of the countless challenges, complications, administrative inefficiencies, federal self-sabotage, and bureaucratic nightmares, the process also involves workers on the ground who are deeply committed to the mission.

As the days went on, the success stories rolled in. One part of our census area went from a 16 percent completion rate to 66 percent in two weeks. The Macon area, which had the second- or third-worst response rates out of 248 census areas across the country, started getting national recognition in leadership calls. Other team members would be dispatched to Alabama and Louisiana, while travel assignments also came up in Illinois and the tribal areas of Arizona. Back in Yonkers, my home district was training enumerators to visit homeless shelters and food banks.

But for all these small wins, the census was no match for 2020. During shifts in Georgia, it hadn’t been unusual to receive dirty looks or stray comments about wearing a mask, even as some residents warned us that people on their block had COVID-19. More devastating yet, enumerators in Rochester, New York, were being sent out into the field in pairs after heated protests broke out following reports that Daniel Prude, a mentally distressed Black man, had died of suffocation while being taken into police custody, a development that added the city to the ranks of the many to experience unrest over issues of racial injustice this year.

The months-long delay of the counting period also placed the work in the middle of natural-disaster season. While wildfires were blazing in the West, threatening basic life in California, Oregon, and Washington, much of my team sat idle for a day as Hurricane Sally dumped heavy rain on Georgia.

These sorts of delays also mean that the final census results will now be assembled in a heavily compressed time frame. The Trump administration originally asked Congress to give the Census Bureau until the end of April 2021 to process and check this decade’s data. But this August, administration officials changed their mind, and said they could get the work done by December 31, 2020. As a result, census administrators are now trying and (apparently failing) to complete their high-stakes work at twice the normal speed—a fittingly messy end to a particularly chaotic count.

Thinking back, though, the most striking moments of the months I spent working as a census taker were the ones when Americans’ responses to my questions brought out the deep imperfections of the process, and the enduring divisions of the nation. Sometimes, when an American told me they were Hispanic, they’d rush to adamantly assure me that they were a legal U.S. citizen too. Those moments reminded me that, for 20 years, the census counted all East Asian people as “Chinese”; that many censuses counted enslaved Americans as three-fifths of a person; that the census once asked whether a household contained anyone who was “insane” or “idiotic”; that the Roosevelt administration used census data to target Japanese Americans for internment. How we plan, conduct, and use the census reveals the character of the country in a way that numbers alone cannot.

Later this month, the Supreme Court is set to hear arguments about whether President Trump may omit undocumented immigrants from census rolls when apportioning seats in Congress. Many census advocates argue that this would be a violation of the Fourteenth Amendment, which directs “counting the whole number of persons in each State.” But that’s the whole fight: Who counts as a person, anyway?