Tuesday, August 16, 2016

#CensusFail, or, how #Census2016 is not as big a hit in Australia as in Canada

I've learned that the Australian census, recently concluded, has been the subject of as much controversy as in Canada in recent years. The big difference is that, whereas Canada's census controversies were contrived by the then-ruling Conservative Party government, Australia's worries are more deeply rooted.

Bloomberg's Michael Heath authored the article "Census Boycott Gathers Momentum Amid Australia Privacy Concerns".

A backlash against Australia’s national census is gathering momentum with lawmakers joining calls to boycott Tuesday’s population count amid concern data gathered will be used to build wide-ranging profiles of individuals and violate their right to privacy.

Nick Xenophon, who leads a minority party in parliament, is refusing to provide his name to the compulsory census and thus won’t be able to submit a completed document, risking a fine of A$180 ($138) a day. Other lawmakers , including Greens Senator Sarah Hanson-Young, have also threatened to withhold their names.

Fearing a widespread boycott that could compromise data that’s essential for public service provision, the government has sought to downplay the concerns.

“Privacy matters,” Xenophon told reporters. The Australian Bureau of Statistics “has failed to make a compelling case on why names must be provided and stored for four years. All names will be turned into a code that ultimately can be used to identify you.”

The crux of the debate is over a “statistical linkage key” that will be created for an individual from the name submitted on their census form. Names will also be kept for four years rather than being destroyed after 18 months, as is the current practice.

Privacy advocates say that, irrespective of names being destroyed, the linkage keys will allow answers to future questions to be linked to census responses, enabling the government to compile a profile of a person.

Edward Johnson, also writing for Bloomberg, produced "Census Crash Spurs Australia Security Fears in Blow to Turnbull".

Prime Minister Malcolm Turnbull’s government is facing its first big test since scraping back into office last month, after an apparent overseas attack crashed Australia’s online census and threatened to derail the survey.

The five-yearly census, which is used to underpin economic planning and public service provision, was shut down late yesterday after four so-called denial of service attacks, according to the Australian Bureau of Statistics. The failure is embarrassing for the government, which has defended the integrity of the survey amid calls for a boycott over privacy concerns.

The statistics bureau shut down the site out of an “abundance of caution” to ensure the data already submitted by some 2 million Australians couldn’t be compromised, Treasurer Scott Morrison said Wednesday at a news conference with Turnbull. He joined the prime minister in urging Australians to complete the survey once the site was restored, saying it was “critical to support economic planning.”

This year’s census had already been mired in controversy as lawmakers joined calls to boycott the population count amid concern data gathered could be used to build wide-ranging profiles of individuals and violate their privacy. The debacle is an unwelcome distraction for Turnbull, who was returned to office with a wafer-thin majority of just one seat after the July 2 election -- a result that has eroded confidence in his leadership.

“The sense of trust and goodwill toward the census has been tarnished,” said Zareh Ghazarian, a politics lecturer at Monash University’s School of Social Sciences. There’s a risk that people will “have greater reservations now about completing the survey” and the data garnered will be incomplete, he said.

I have seen Australian friends actively complaining on social networking sites about their problems filing their electronic returns.

Do any readers know more about this? If you are Australian, can you explain your perspective? I'm quite curious, and I'm sure our other readers are, too.

Saturday, August 06, 2016

On the Jedi phenomenon and the Australian census

Via io9, I learned of a Brisbane Times article dealing with the Jedi phenomenon in the Australian census. Funny as the idea of claiming the faith of the Jedi Knights of Star Wars fame might be, it messes with census data.

The Jedi phenomenon began in 2001 when an email campaign mistakenly claimed the government would have to recognise it as an official religion if 8000 people selected it in the census.

Now [HKlie] Sturgess, president of the Atheist Foundation of Australia, is leading a campaign for people not to treat the census as a joke.

This is because if people fill in the "other" box in the religion section of the census with an answer such as Jedi they are counted as "not defined" rather than "no religion".

Ms Sturgess said this skews the census results by making Australia appear more religious than it is.

"People shouldn't waste their answer," she said.

"Answering the religion question thoughtfully and honestly matters because it benefits all Australians when decisions on how to spend taxpayer dollars are made on sound data that accurately reflects modern-day Australia."

This little incident says a lot about the problem with modern statistics. I will elaborate in the next few days. In the meantime, be amused with this.

Tuesday, August 02, 2016

Four article links on the aftermath of the 2016 Canada census

What, might readers of Demography Matters ask, has become of the Canadian census? Four article illustrate the contours surrounding this issue.

Last week in The Globe and Mail, Tavia Grant's article "Census response rate is 98 per cent, early calculations show" featured.

Canadians really were, it seems, enthusiastic about the census.

Statistics Canada is still calculating exact response rates, but it says early indications are that the overall response rate is 98 per cent – and about 96 per cent for the long-form census. That is higher than long-form response rates in the previous two censuses, the agency says.

“Early indications are positive,” Marc Hamel, director-general of the census program, said in an interview.

These numbers could shift up or down as results from early enumeration of Northern communities, late filers and First Nations reserves are added in, he said. “The range of error is not very high … it’s likely to move, but we’re talking most likely, at most, one percentage point.”

[. . .]

The sample size for the long-form census was increased to one in four households this year from one in five in 2006. The combination of high response rates this year and a bigger sample size will yield “incredibly precise data,” chief statistician Wayne Smith said.

Grant also had another article in The Globe and Mail, "Statistics Canada's tech issues hampering its mandate: chief".

Statistics Canada’s technological troubles have become so acute that its chief statistician says they are hampering the agency’s ability to carry out its mandate – and he places the blame squarely on one source: Shared Services Canada, the department now running the agency’s informatics infrastructure.

Statscan’s website has for months been beset by crashes, delays and outages, most notably on July 8, when its main website was down for more than seven hours on the day of the release of the labour force survey.

In an interview with The Globe and Mail at his Ottawa office, chief statistician Wayne Smith said that that outage – along with a long list of other information technology troubles – relates to problems with Shared Services.

“It’s had a significant impact on our operations,” Mr. Smith said. “Our service to the public has suffered, clearly, in ways that we would rather not have happened. Some of our relationships have suffered. … There’s a frustration among our clients.”

Canada’s statistical agency is tasked with producing quality data and analysis about the country on everything from oil exports to jobless rates, food prices and health outcomes. That mandate, Mr. Smith said, is at risk as tech glitches – stemming partly from a lack of maintenance at its data centre – have caused delayed releases, lost time in conducting quality assurance and higher costs.

The CBC carried Jordan Press' Canadian Press article "StatsCan looking for powers to make all surveys mandatory, compel data from companies".

Statistics Canada is privately floating the idea of new powers that would make all of its surveys mandatory by default and force certain companies to hand over requested data, such as credit card transactions and Internet search records.

Currently, the agency can ask for any information held by governments and businesses, but officials have long found it hard to get information like point-of-sale transactions that could give a more detailed and accurate picture of household spending.

The agency's proposal would compel governments and companies to hand over information, and levy fines to discourage "unreasonable impositions" that "restrict or prevent the flow of information for statistical purposes."

Corporate fines would depend on a company's size and the length of any delays. The changes would also do away with the threat of jail time for anyone who refuses to fill out a mandatory survey, such as the long-form census.

The recommendations, contained in a discussion paper Statistics Canada provided to The Canadian Press, would enshrine in law the agency's independence in deciding what data it needs and how to collect it.

In the National Post, John Robson's "When statisticians corrupt" counsels against this extension.

Lord Acton famously warned that “Power tends to corrupt, and absolute power corrupts absolutely.” Even, it turns out, Statistics Canada.

Now you laugh. But my colleague Kevin Libin just wrote about how Canada’s statistical agency, giddy with success in having the mandatory long-form census restored, now wants all sorts of power to compel you to reveal your intimate secrets or lack of same.

I don’t want to be seen mocking statisticians as good at math but without the personality to be accountants. Especially not now. But it is hard to imagine a less menacing bunch than the geeks at Statistics Canada, whom I have always found obliging when seeking data on deadline. They wield spreadsheets in cubicles, not pistols in dank secret police HQ basements. Yet there they are, demanding the power to compel compliance, avoid scrutiny, even run their own computers to avoid depending on their wretched colleagues for tech support.

‘Twas ever thus in the executive branch. We know best (as Libin notes, Statistics Canada has elevated itself from a good number cruncher to “a key institution in the democratic process”) and should be freed from petty restraints on our capacity to act for the greater good. But why?

Well, in Acton’s and my vision of the political problem, we need the state and it needs extensive powers to protect our freedom. But anyone to whom we grant extensive powers is liable to intrude on our freedom because, as Alexander Solzhenitsyn wrote, “The line separating good and evil passes not through states, nor between classes, nor between political parties either, but right through every human heart.”

That is where Canada is right now.