Tuesday 9 May was a busy news day. In the space of a few hours, journalists had two major stories to contend with: the firing of FBI director James Comey, and the retirement of US Census Bureau director John Thompson.
Comey may have got more headlines, but Thompson’s departure was “arguably as consequential to the future of our democracy,” wrote Time magazine. “That’s because the Census Bureau … plays a staggeringly important role in both US elections and an array of state and federal government functions.”
Thompson’s retirement was seen as coming at a critical time for the Bureau – just three years away from the decennial Census, in which every individual within the United States is counted. The Census is mandated by the US Constitution – Article 1, Section 2 requires an “enumeration” to take place every 10 years – however, many policy-makers have become frustrated by the increasing costs of the Census, from $3 billion in 1980 to $12.3 billion in 2010 (dollar amounts are given in 2020 constant dollars).
When Thompson was sworn in as director in August 2013, he had in mind the goal of holding costs stable. The Bureau’s own estimate was that a 2020 Census, conducted the traditional way, would require $17.8 billion. Instead, Thompson and his team proposed to change the Census and keep costs at $12.5 billion.
However, innovation requires investment, and Census watchers worry that, to date, the Bureau has not received the money it has asked for, and needs, to put in place new systems and processes while still delivering an accurate count of the US population.
Just days after his official departure date of 30 June, Thompson spoke to Significance about his time at the Bureau, his ambitions and achievements, and the challenges ahead for the 2020 Census.
Time to modernise
Thompson was no stranger to the Census Bureau when he was nominated by President Barack Obama to become its next director, following Robert Groves. Thompson had previously worked for the Bureau from 1975 to 2002, leaving to join NORC at the University of Chicago, a research institution of which he became president and chief executive in 2008.
“I was very pleased to go back [to the Bureau],” he says. “I knew a lot of the people that were still there. I had been on the Committee on National Statistics and had served on two National Academy panels that were giving advice to the Census Bureau, and so I really had in mind what I wanted to do at the Census Bureau from that experience.”
Thompson had two key ambitions: to modernise the way the decennial Census was conducted, and to push the Bureau to look at alternative ways of collecting information, including the use of administrative data. Survey-based methods of data collection have been honed and improved since the turn of the century, Thompson says, “but they are really being stretched right now”.
Consider the Current Population Survey, which provides monthly estimates of unemployment in the USA. “You can see that from 2004 its response rates were fairly level until 2010, when they got kind of a bump [up] because of all the advertising for the 2010 Census,” says Thompson. “But since then the response rates have been steadily dropping off and the costs have been going up, and the same is true for a lot of our other surveys, particularly some of our economic surveys where the response rates are really going down. It’s becoming essential that we find alternative ways to collect information that don’t rely as heavily on surveys.”
Thompson says that there are research programmes under way to look at how administrative data can be used to enhance, for example, the American Community Survey, a yearly survey looking at jobs and occupations, education, home ownership and other aspects of the US population. He also refers to “an active research programme in the economic directorate to look at transaction data to improve retail statistics”.
But some of the most ambitious – and potentially risky – innovations are proposed for the 2020 Census.
To hear Thompson tell it, the decennial Census is almost unchanged from a design that was first put in place for the 1970 Census. It involves compiling a nationwide address list, assigning geographic codes, mailing out surveys that are then mailed back, and capturing information from the completed surveys using optical scanners. But of course not everyone sent a Census form sends one back, and so the Bureau must hire an army of fieldworkers, armed with paper and pencil, to go knocking on doors to collect the missing information.
Thompson says that sticking to this design while the US population has become more diverse, particularly in the 1990s and 2000s, explains the rise in costs. With a paper-and-pencil operation, he says, “the only way you can deal with that increase in complexity and diversity in living arrangements is to put more people on it, and that just increases the costs dramatically”.
For 2020, Thompson and the Bureau proposed four major changes. First, the internet would become the primary self-response option, rather than mail. Second, the Bureau would use advances in geospatial data and satellite imagery to update its address list. Third, it would equip fieldworkers with mobile devices to collect information from those individuals who do not self-respond. And fourth, the Bureau would use administrative data to identify and remove vacant housing units from the universe of non-responding households; then, after visiting every occupied non-responding household at least once, it would use administrative data to count the remaining non-responding households, thus reducing the need for fieldworkers to make repeat visits.
According to the Census Bureau: “Together, these innovations are expected to result in a Census that costs $5.2 billion less than it would have cost if the 2010 Census design were repeated in 2020.”
But it is Census funding – rather than cost savings – that have been of most concern of late. Advocates, including those at The Census Project, worry that the allocated funding is not sufficient to ensure a successful, accurate count. The Bureau spent much of fiscal year (FY) 2017 operating on a “continuing resolution”, which meant that money for the Census was held at the same level as FY 2016 – around $600 million. Planned activities had to be deferred to later years, “including some of the address list updating and, more importantly, some of the work on advertising and partnership, which is really critical to producing an accurate Census”, says Thompson.
Partnership work has been “incredibly important in the past couple of Censuses”, he says. It involves hiring people to go out and work with local governments and local communities to explain why it is important to respond to the Census, and how the information that is provided is kept confidential. “That’s what generates the accuracy for the Census, and if that doesn’t get set up, I would agree that the accuracy of the 2020 Census will be at risk,” says Thompson.
However, he believes there is sufficient money for the 2020 Census in the FY 2018 budget – $800 million – to develop systems on “the critical path to the 2018 end-to-end test”, a mini Census that is scheduled to take place in Providence, Rhode Island. “This will test not just the effectiveness of individual systems, but how they interface together and pass data back and forth,” explains Thompson.
“Right now, it’s time to be concerned. It’s not time to panic,” he says. “But FY 2019 is going to be an incredibly important year … just incredibly important.” In 2019, nothing else can be deferred. “You have to really finish up the systems that you need to finish … [and] start laying out the infrastructure you need to both collect the Census and promote the Census locally so that the population will respond appropriately,” says Thompson.
We asked whether funding concerns of this sort were a factor in his early departure, six months before his term was due to end. “No,” says Thompson. “You reach a point where you are doing things and you have done everything you can do, and you can’t do any more. In fact, sometimes it is counterproductive to continue to try to do that. It was just at the point where I couldn’t do any more for the Census Bureau and it was time to let someone else do [the job].”
The day before Thompson’s retirement, Secretary of Commerce Wilbur Ross announced that Ron Jarmin, associate director for economic programmes, was to become interim director of the Bureau, while Enrique Lamas, associate director for demographic programmes, would become interim deputy director.
Thompson views the appointments positively, calling them “a strong indication that the Department of Commerce and the [Trump] administration [have] confidence in the career people at the Bureau”. But with the clock ticking on the 2020 Census, it would not do to leave the Census Bureau leaderless. It needs support “in just about all the areas to develop the systems and to advocate for the funding, particularly in FY 2019, to put the infrastructure in place to get an accurate count”, says Thompson.
- This interview is taken from our upcoming August 2017 print edition. Subscribe here.
- As Significance was preparing to go to press, we received news, via press release, that John Thompson had been appointed executive director of the Council of Professional Associations on Federal Statistics.