Census 2020: Armed with the iPhone 8, canvassers are going modern

Update, Aug. 4: After starting nonresponse followups in limited areas in July, the US Census Bureau announced on Monday that it will cease all field data collection by Sept. 30. That's a month earlier than originally planned. As of Aug. 3, only 63% of the country's households have responded to the census. In comparison, the 2010 Census had a final national response rate of 74%. Our story on the 2020 census, which originally published April 1, follows below.

Michael Thieme, the man who oversees the US Census Bureau's software and IT teams, knows the bureau has a target on its back. This year's national decennial census, the first of which began after the Revolutionary War, is plunging full-force into the 21st century thanks to a pricey upgrade project. In the next few months, thousands of bureau employees, known specifically as enumerators, will fan out across the country to conduct one of the federal government's largest peacetime efforts. And for the first time ever, instead of using a clipboard and paper, they'll be counting people with an iPhone app.   

"We certainly don't have any hubris," said Thieme, the assistant director for decennial census programs, systems and contracts. "We had to build something that would handle something that's never been handled before."

The idea of a census sounds deceptively simple: Just count everybody in the country and count them in the right place. But pulling off a complete and accurate count of the estimated 300 million people living in about 140 million households is an enormous undertaking. While the bureau is mandated to do this by the Constitution every 10 years (and designated April 1, 2020 as Census Day, its national day of observation), this year's efforts are further complicated by the coronavirus pandemic, which has locked down the vast majority of Americans in their homes.   

That won't stop the enumerators and their critical task. The census does far more than establish redistricting boundaries or determine the number of seats a state receives in the House of Representatives. It's also used to help companies like Target decide whether to build another location in your neighborhood, if a new freeway onramp should open up along your commute, or if your local senior center or homeless shelter receives enough federal funding for the year. In perhaps the most pertinent example of the census' importance as the nation grapples with COVID-19, its data also helps allocate emergency readiness needs.  

"Every district from the smallest school board to congressional districts depends on accurate census data," said Stanford Law School professor and census and redistricting expert Nathaniel Persily. "It is critical to any input of any sensible public policy."

In an effort to make the door-to-door process, which is the most laborious and expensive part of the census, faster and more efficient, the bureau is arming 500,000 enumerators with the Apple iPhone 8 . But as the census goes mobile, instantaneously beaming respondents' answers to data centers and cloud servers, it opens itself up to those who may want to access or manipulate such valuable information. The stakes to pull off a census have always been high, but with this year's adoption of new technological methods, the pressure to succeed is even higher.

A long and tumultuous history

With its $5 billion total IT investment to usher the census into the digital age, the bureau actually hopes that most residents will not meet an iPhone 8-wielding canvasser. Instead, it wants the majority of people to respond via an online portal, which has been available since mid-March. The portal was primarily built and is maintained by Pegasystems, a Massachusetts-based software company, who also designed the app enumerators will use. 

People can also complete the census over the phone or mail. But for households who don't answer, enumerators will be sent out starting May 28 for what's called a nonresponse followup. (Due to the coronavirus, the bureau pushed its original date from May 13.)

The effort to give canvassers a handheld computing device has a long and tumultuous history that reaches back to the 2010 census. In 2006, the bureau awarded defense contractor Harris Corporation a $600 million contract to provide 600,000 PDAs. Custom-built by HTC and known as the HTC Census, the blue and gray plastic device had a small display that sat above a set of physical keys. It had GPS and was successfully used to track housing coordinates and log addresses of existing and new buildings.

The bureau also planned for the PDAs to be used for nonresponse followups, but the venture fell apart in the end. According to a July 2011 report of the management challenges of the 2010 census by the bureau's then senior research scientist, Daniel Weinberg, the project failed partly because of a lack of rigorous software testing and early underestimates of the cost and time the endeavor would require.

In 2008, the bureau decided to abandon the project and revert back to paper forms. By then Harris Corporations' contract ballooned to a total of $1.3 billion, though the overall cost of the 2010 census came in under budget. The fiasco was "one of the most expensive failed software systems in history," according to a report by the University of Minnesota's Minnesota Population Center.

For the 2020 census, which is estimated to cost a total of $15.6 billion, the bureau has to avoid making the same blunder again.

Weighing the competition

When the bureau began to consider pitches from potential contractors in 2015, many devices for data collecting out in the field were suggested, including Android phones and tablets . After weighing the competition, the bureau went with CDW-G in June 2017, a tech company headquartered in Illinois, who pitched the leasing of iPhones.

The iPhone attracted the bureau in two key ways. First was familiarity. Apple is one of the most popular phone-makers in the world, and many of the canvassers, who undergo seven to eight weeks of training and are paid an average of $20 an hour, would already be familiar with its iOS software. CDW-G initially proposed using the iPhone 6 , but once the iPhone 8 went on sale in 2017, the bureau was able to upgrade to it without an increase in cost. 

The iPhone is stable and its operating system has foundational protections that, while not unique to iOS, are still attractive to enterprises. One is code signing, a feature that prevents apps from being recognized if they alter their code after installation. Another is sandboxing, which restricts apps so they can't "reach" outside and access other apps. 

Because these features are available on all iPhones, the models used by the bureau don't need to be custom modified. So if an enumerator's iPhone breaks, for example, they could theoretically walk into an Apple retail store, buy an iPhone and be up and running again after enrolling the new iPhone into the census' automated mobile device management system. (Though in real life, this would not be protocol.)

"We are proud that the US Census Bureau will use the iPhone for the collection and management of 2020 census data," said an Apple spokesperson. "The census is an important constitutional cornerstone that aims to ensure every one of us has equal representation, education and access in America."

To the benefit of CDW-G specifically, iPhones have high residual value compared to its competitors too. After the door-to-door phase is scheduled to wrap up on Aug. 14, the bureau ships back the iPhones to CDW-G, which wipes them and removes them from the census' system. They are then repurposed or resold by CDW-G for other enterprise purposes.

The movie plot scenario

Pegasystems had to tackle a certain set of issues that are unique to the census in the app's design process. The app had to be user-friendly, given that hundreds of thousands of temporary workers from different backgrounds were going to use it. It had to work on any type of phone -- development of the app began before the bureau ultimately decided on the iPhone. And the app had to work with or without a network connection given the variety of locations canvassers will travel. 

"The app was designed based on the assumption that there would be no connectivity," said Lisa Pintchman, Pegasystems' vice president of corporate communications. "This app is designed to be used and work exactly the same on the ground floor of a New York City high-rise or a remote town in North Dakota."

On top of entering census information from respondents in the app, workers can also use it to read their prepared script, check their case list, log their hours and expenses and connect to the help center if they run into any issues. At the end of every interaction with a respondent (or attempted interaction), workers can type in post-interview notes to pass along in case a second followup is necessary. This can be about anything, from suggesting that the next canvasser to revisit be fluent in Russian, or that there's a big dog in the front yard that colleagues should be wary about.

So far, results appear promising. In a sort of "dress rehearsal" for the 2020 census, which was conducted in 2018 in Rhode Island, enumerators completed 1.56 cases per hour worked using the app, compared to 1.05 cases during the 2010 census. In a 2018 release, the bureau called the 49% increase in productivity a "remarkable improvement."

But as with any attempt to digitize such a huge venture, using an app brings its own security dangers. 

"The movie plot scenario is that this information is accessed," said Stephen Moore, chief security strategist at Exabeam, a cybersecurity company located in California. "That'll erode the competence of the system and it'll decrease participation, which may be the goal of some foreign adversary, and that's a risk."

For example, the time when your census information is most vulnerable, according to Moore, is while the enumerators type in the data and right when they submit your answers (after submission, once census data has been delivered to the bureau's servers, it is automatically deleted locally from the device). 

To minimize this danger, the bureau implemented many security measures, including multifactor authentication, the use of security tokens and end-to-end encryption.

"The data that [enumerators] collect does not stay on the device very long," said Thieme. "Our application constantly uploads data to our backend, and the data is encrypted on the application, it's encrypted in transit and it's encrypted at rest when you give it to the Census." 

The iPhones' features have also been limited by the bureau and the devices can only be used as a census tool. Though they can make calls, workers can't browse the internet or download additional apps from the App Store. If a phone is reported stolen or lost, it will be remotely bricked and completely nonfunctional. 

The bureau also distributed its online footprint to servers all over the country to prevent a potential flooding or overwhelming of its app (this would avoid a similar situation to the Iowa Caucuses in February) and would help thwart a malicious denial-of-service attack (which happened during the 2016 Australian Census).

Nevertheless, nothing is 100% impenetrable. As recently as February, a report from the Government Accountability Office recommends the bureau undergo even more testing and assessments "to improve system performance and scalability." But to the bureau, the benefits of a modern census outweigh the risks. In addition to working alongside the FBI, the Department of Homeland Security and other federal agencies, the bureau itself knows it must remain hypervigilant against nefarious actors, be it foreign or domestic, who either want to access sensitive information or alter it to gain political advantages. 

"Our security group is always in high gear, but it has never been in higher gear than now," said Thieme. "It's a dangerous world … all we can do is watch it and be ready to react."