PHOTO: CRAIG F. WALKER/THE BOSTON GLOBE/GETTY IMAGES564 6 MAY 2022 • VOL 376 ISSUE 6593 science.org SCIENCEadopt meaningful improvements in today’s
hyperpartisan political climate.
The agency’s constitutional duty—to
determine how to divvy up the 435 seats in
the U.S. House of Representatives by state—
makes it a perennial target for politicians.
The 2020 cycle was especially fraught as
then-President Donald Trump waged an as-
sault on the agency’s independence that in-
cluded an unsuccessful campaign to add a
citizenship question to the census.
One major challenge for the agency is
to reduce its reliance on a master list of
addresses, painstakingly compiled and
continuously updated, as the basis for the
decennial head count. But counting every
household requires an army of enumerators
to make repeated follow-up visits to dwell-
ings believed to house residents who did not
respond to the initial census questionnaire.
The Census Bureau hopes to dramatically
shrink that mop-up effort in 2030 by iden-
tifying most of the U.S. population through
so-called administrative records: personal
information residents have already pro-
vided to other federal agencies, as well as
data collected by local and state govern-
ments and by commercial vendors. The
bureau could then conduct a much smaller
follow-up effort to track down the estimated
15% to 20% of the population lacking a us-
able electronic footprint.
There’s no guarantee, however, that data
culled from administrative records would
meet the bureau’s high standards. And enu-
merating people who are missing from those
records is no trivial matter, says Joseph Hotz,
a population economist at Duke University.
Hotz is conducting a study to understand
rural aging in North Carolina and has found
that men living alone can be especially elu-sive. “They aren’t homeowners, and they
don’t appear in commercial housing data-
bases,” he says. “They aren’t easy to find be-
cause they don’t want to be found.”
Another major goal is to make better use
of data from the welter of surveys the bureau
now conducts apart from the decennial cen-
sus. Those surveys generate separate master
lists, or frames, that contain data such as the
number and location of housing units, a ros-
ter of employers and jobs, and demographic
information for the population as a whole.
Combining those frames and analyzing
the merged data, Santos says, “might answer
policy questions you couldn’t even put on the
table before.” It could also lead to shorter, less
burdensome surveys, he adds, by eliminating
redundant questions.
The UVA team is conducting a series of pi-
lot studies to demonstrate the added value of
what it calls a “curated data enterprise.” Joe
Salvo, former chief demographer for New
York City and now a senior fellow at UVA,
says one example of an important economic
trend that current surveys have struggled to
capture is the gig economy, in which people
hold down multiple jobs.
The bureau’s Current Population Survey
( C P S ) , a m o n t h l y s u r v e y o f t h e U. S. l a b o r f o r c e ,
yields a gross undercount of gig workers by
assuming respondents have just one primary
job, Salvo asserts. “CPS says there are 15 to
18 million people in the gig economy, and
that the number hasn’t changed since 2005,”
he says. “But recent studies suggest the num-
ber could be as high as 55 million.” He thinks
a better answer would come from combining
CPS data with Internal Revenue Service data
on reported income, along with records from
nonprofit organizations and from those who
are self-employed.Another UVA project, on the state of nurs-
ing homes and residential care facilities, aims
to create a more detailed picture of those who
live in, work at, and operate such facilities.
Keller says that picture could be pieced to-
gether from geospatial data and business and
labor surveys and would be invaluable for re-
sponding to a natural disaster or pandemic.
Despite the potentially huge payoff, such
efforts raise the specter of the federal gov-
ernment as an all-knowing Big Brother.
Some of those fears could be alleviated,
Keller says, by pooling the data only as
needed, to answer a specific policy question.
Another safeguard is to cloak personal,
individualized data. “We’re very close to
having the technology for creating a fire-
wall that allows access to the data without
actually being able to look at individual re-
cords,” Santos explains.
The Census Bureau is already prohibited
from releasing any data sets that, when
combined with other anonymous informa-
tion, could allow outsiders to identify indi-
viduals. To meet that requirement, the 2020
census applied a mathematical tool called
differential privacy, which injects statistical
“noise” into the data.
However, many demographers have com-
plained that the amount of noise needed to
avoid disclosure has undermined data qual-
ity. They worry the problem will get worse
if the Census Bureau sets too high a privacy
bar for any data it releases. One proposed
solution would be to exempt some data,
such as addresses, that are typically avail-
able on the internet.
Privacy concerns aren’t the only reason for
the decline in survey participation, social sci-
entists say. The public is also skeptical that
surveys are worth the time. “It’s insufficient
to tell people they are required by law to par-
ticipate,” Santos says. “We need to demon-
strate tangibly how the data we’re collecting
is benefiting them on a day-to-day basis.”
Nancy Potok, former chief statistician
for the United States and a former deputy
Census director, wholeheartedly endorses
what Santos and the UVA team are trying to
accomplish. But she wonders whether the
agency will be able to make headway.
“Any big changes will make the Census
Bureau a [political] lightning rod,” she pre-
dicts. “And Census tries hard to avoid that
fate. So in the end, I think that the status
quo ... is the most likely outcome.”
Prewitt hopes otherwise and is count-
ing on the research community to help
the agency make its case for change. “Our
challenge is to defend something that’s in-
herently political with something that isn’t
political,” he argues. “And the answer is sci-
ence. That’s the only thing that will carry
the day.” jExisting U.S. surveys provide an incomplete picture of elderly care facilities.NEWS | IN DEPTH0506NewsInDepth_15558970.indd 564 5/3/22 5:45 PM