America’s census looks out of date in the age of big data
Cost, complexity and new sources of data mean the days of the population count are numbered
A DOG-SLED or a snowmobile is the surest way to reach Toksook Bay in rural Alaska, where Steven Dillingham, the director of America’s census bureau, will arrive to count the first people in the country’s decennial population survey on January 21st. The task should not take long—there were only 590 villagers at the last count, in 2010—but it marks the beginning of a colossal undertaking. Everyone living in America will be asked about their age, sex, ethnicity and residence over the coming months (and some will be asked much more besides).
This census has already proved unusually incendiary. An attempt by President Donald Trump to include a question on citizenship, which might have discouraged undocumented immigrants from responding, was thwarted by the Supreme Court. His administration has also been accused in two lawsuits of underfunding the census, thus increasing the likelihood that minorities and vulnerable people, such as the homeless, will be miscounted.
America’s constitution mandates that a census take place every decade so that legislators “might rest their arguments on facts”, as James Madison put it in 1790. The government has become more reliant on this knowledge as its responsibilities have grown. In 2016 census data were used to direct some $850bn of funding for programmes such as Medicaid, food stamps, school lunches and roadbuilding. The results are also used to apportion seats in Congress, as well as by academics, genealogists and even supermarket chains deciding where to open new shops.
Population counts long predate the founding fathers. Babylonians recorded their numbers on clay tiles as far back as 3800BC to work out how much food to grow. In ancient Athens administrators counted piles of stones, one added by each citizen, to gauge military capability and tax revenues. And Joseph and his pregnant wife Mary travelled from Nazareth to Bethlehem after Emperor Augustus decreed that “all the world should be registered”. By the 18th century, reliable and regular population counts were common in European countries, and enumerators (as census-takers are known) were being sent out to colonies around the world.
A decennial survey of every household, as has just begun in America, is a tried and tested method. It provides a snapshot of an entire population. Citizens can state how they wish to be recorded and the resulting treasure-trove of data is publicly accessible. But the cost and scale of such an undertaking is growing. America’s previous census cost $92 per household, up from $16 in 1970 (in 2020 dollars). China mobilised an army of 6m enumerators to roam the country in 2010. The UN Population Fund calls a census “among the most complex and massive peacetime exercises a nation undertakes”.
Migration and changing lifestyles are making it more difficult to reach everyone. Renters are trickier than homeowners to count reliably, because they move more often and live in less stable households. One study projected that this year’s American census could undercount the population by 1.2%, rising to more than 3.5% among black and Latino populations, who are less likely to own their home.
Is there a better way? For the first time this year, Americans will be able to fill out the census online. This risks missing hard-to-reach groups such as indigenous populations and the old. It also introduces unforeseen headaches. In 2016 Australia’s census website crashed, leaving millions unable to submit their responses and venting their anger with the hashtag #censusfail.
Nordic countries have ditched the unwieldy undertaking altogether, turning to other sources of information. In Sweden each citizen is given a personnummer, an identity number linked to government data on individuals’ health, employment, residence and more. These data are cross-referenced to produce statistics resembling the results of a traditional census. Denmark, Finland and Norway take the same approach. As societies share more information, wittingly or otherwise, new statistics can be produced. Mobile-phone records, for example, have been used to estimate commuting patterns. The Netherlands, meanwhile, conducts what it calls a “virtual” census. This is similar to the Nordic model, but also uses small-sample surveys to produce data not already held by the state, such as education levels and occupation.
As long as each citizen has a unique identifier, such counts are cheaper to carry out—the Dutch government boasts that its census in 2011 cost just $0.10 per person—and can be done much more regularly. But the accuracy of the data is harder to guarantee. Population registers are never completely up to date and anyone not already on them will be missed. In Europe, two-thirds of countries are expected to use data from existing registers to some extent in the next round of censuses. This is up from just a quarter 20 years ago, according to analysis by Paolo Valente, a statistician at the UN.
Making such a change is a slow process. Bernard Baffour, a researcher at the Australian National University, points out that it took decades for Sweden to implement a fully register-based census, partly because Swedes had to be reassured that their data were secure. As he puts it, “When a doctor asks how much you drink or smoke, are you happy for that to be linked with all the other information on you?” Frank de Zwart, a professor at Leiden University in the Netherlands, also criticises register-based censuses for neglecting a key political function of censuses. For minorities such as native Americans, filling out a census is a powerful assertion of their place in society. A virtual census would deny them this opportunity. That said, self-reporting is far from perfect: 177,000 Britons implausibly claimed to be Jedi knights in the census of 2011.
Even though Britain does not have identity cards, common in the rest of Europe, in 2013 the government tried to replace the census with other administrative data it already held. An outcry from MPs and statisticians forced ministers to shelve the idea. The public had rejected an attempt in 2006 to introduce identity cards, and recent scandals such as the harvesting of personal data from Facebook deepened Britons’ worries about privacy. Iain Bell, the statistician in charge of the census at the Office for National Statistics (ONS), emphasises the importance of public trust in producing official figures: “If people don’t want a single register of the population, we have to respect that and look to other sources.” Francis Maude, then a government minister, told MPs in 2014 that he hoped the next census, due to take place next year, would be the last. In 2023, the ONS will report back on whether this is achievable.
Political rows over America’s census have shone a light on a function of government that most people consider only a handful of times over their lives, but the results of which affect them every day. Recording each member of every household seems outdated in the age of big data, whether the data are held by governments or private companies. But in this sense, at least, America’s federal government is not big enough; its social-security system is too incomplete, and other information still too patchy, to replace the old-fashioned head-count. Will Mr Dillingham be the last enumerator to visit Toksook Bay? Don’t count on it.
No comments:
Post a Comment