How the census went from a quinquennial chore to a national crisis -

How the census went from a quinquennial chore to a national crisis

No fun with numbers


Rick Eglinton/TORONTO STAR

“Wrong,” Tony Clement typed to an entity calling itself Harbles. “Statisticians can ensure validity w larger sample size.” A week into the Great Census Uprising of 2010, the industry minister had taken to Twitter, as he often otherwise does to detail both his travels and his music tastes, in hopes of making the case for changes to the national head count.

“Wrong,” countered Stephen Gordon, an economist at the University of Laval. “Large samples can’t fix sample selection biases.” Clement tried again, tweeting that “proper weighting” would be used. “Where will the weights come from?” Gordon shot back. “Other voluntary surveys get their weights from the census.” After another exchange of tweets, the minister fell silent for the night.

This was, in its odd way, the defining moment of what has become a profound debate over paperwork and a remarkable outbreak of painstaking seriousness in our political discourse. Where once the census was merely a quinquennial formality, it is now the source of the summer’s pre-eminent political debate, drawing economists, city planners, statisticians, minority groups and religious leaders into a battle that goes to the very purpose and practice of government.

The national census, conducted every 10 years between 1870 and 1951 and every five years since, has recently included two forms: a brief set of basic questions sent to 80 per cent of households and a more detailed long form sent to 20 per cent of homes. In late June, it was quietly announced the long form would no longer be mandatory. Instead, a voluntary survey would be distributed to 33 per cent of Canadian households—the increase apparently meant to offset the fact that a response was no longer required by law. The change made news on June 29 and the next day, Gordon, writing on his economics blog Worthwhile Canadian Initiative, was among the first to proclaim outrage. “It’s not often that sample selection bias becomes an issue of national importance,” he wrote, “but then again, it’s not often that census sampling design is outsourced to drunken monkeys.”

Gordon’s concern was technical, but crucial: response rates to voluntary surveys typically vary among income and education levels, thus skewing the results. Such data can be adjusted, but the usual benchmark upon which that adjustment is based is the census. If its methodolgy is not sound, it’s not reliable, and all the other studies and surveys that depend on it are in trouble too.

From this dilemma followed a storming of the statistical Bastille. Three weeks later, the disgruntled include the Federation of Canadian Municipalities, the Inuit Tapiriit Kanatami, the Canadian Marketing Association, the Canadian Association of University Teachers, the Canadian Federation of Francophone and Acadian Communities, the Canadian Medical Association Journal, the Canadian Jewish Congress, and even the Evangelical Fellowship of Canada. This week, a number of interested observers—including the chief economist at TD Bank and the presidents of the Canadian Labour Congress, United Way and Toronto Board of Trade—wrote to Clement in hopes of meeting to discuss the “considerable economic and social costs” of his decision.

As varied as the complainants are—so far including government officials in at least three provinces—the complaint is essentially the same: anything that undermines the collection of census data undermines the creation and analysis of public policy and the ability of these groups to serve their various communities. “That’s what the Conservatives’ endgame is here—to permanently hobble the government’s ability to enforce legislation and deliver social programs aimed at our most vulnerable,” Liberal industry critic Marc Garneau said last week, suggesting that the unemployed, women and minorities could be most affected.

The government rejoinder is equally as dramatic: that the census in its previous form represented nothing less than an authoritarian intrusion into the private lives of Canadians.

Among other details, the long form asks how many bedrooms are in each respondent’s dwelling, while the Statistics Act includes the threat of a fine (of up to $500) or imprisonment (for as much as three months) for refusing to fill out the census. “Data is valuable to many,” Clement tweeted to Gordon this week. “But personal questions you would like to force Cdns to answer on pain of jail is just plain wrong.”

Statistics Canada, which the minister has said signed off on the decision, has largely declined to defend the new system. And while the mandatory long form’s prominent supporters are multitudinous, Clement’s public supporters have been few. Libertarians have rallied to the government’s defence, but William Robson, president of the conservative-minded C.D. Howe Institute, has lamented their cause. “For those who want governments to do less but do it better, good information is indispensable,” he recently wrote.

Facing such determined and disparate forces, Clement and the Conservatives have turned aggressive. In a memo to reporters last weekend, the Prime Minister’s Office both mocked the census (noting that 21,000 respondents had identified their religion as “Jedi knight” in 2001, the result of a larger prank by Star Wars fans) and attacked the Liberal opposition. “The Ignatieff Liberals promise to force all Canadians to answer personal and intrusive questions about their private lives under threat of jail, fine, or both,” wrote Dimitri Soudas, the Prime Minister’s director of communications.

Apparently now eager to explain themselves, the Conservatives are demanding—echoing an earlier call by the Liberals—that the industry committee be recalled for summer hearings into the matter. Meanwhile, Maxime Bernier, the disgraced cabinet minister reborn as a crusader for classic conservative ideals, has stepped forward as the government’s primary defender. As industry minister during the last census, he claims to have received thousands of emails complaining about the intrusion (though he admits the emails were all subsequently deleted). He also asserts a fundamental governing philosophy. “As I keep saying, government is already much too big and intrusive, and this decision will restore some balance,” he wrote on his blog.

The necessity of the effort in the first place, of course, remains debatable, and not just for statistical reasons. The new process will cost more than the previous census. The 2001 census generated 52 reported cases of non-compliance (the vast majority apparently resolved without trial). According to Clement, after the long form was sent to 2.5 million households in 2006—when the census was subject to protest because Lockheed Martin, the American arms manufacturer, had provided the necessary software—approximately 60 cases were referred for prosecution. Only half a dozen people are reportedly fined after each census and it’s not clear if anyone has ever been imprisoned. Clement claims the government “received complaints about the long-form census from citizens who felt it was an intrusion of their privacy.” The privacy commissioner, though, says the last two censuses generated a total of three grievances.