Thursday, November 25, 2010

A coda for Statistics Canada

The long-form census so vital for Statistics Canada's production of information--something I blogged about this September--is not coming back. Statistical agencies elsewhere in the world have taken note, like the European Statistcal System in its Governing Advisory Board's annual report.

Official statistics have to be professionally independent, strong and of high quality. One indispensable precondition is that the ESS as a whole must be in a position to resist outside pressures when it comes to its core professional competences. Since the peer reviews in 2006-2008, the legislation has been revised to strengthen the professional independence of the statistical services in one third of the countries in the ESS. In five countries the legal process is either under way or being planned. However, legal proceedings can be lengthy and in two countries the revision process is exceedingly slow. Moreover, some of the revisions do not yet guarantee professional independence explicitly enough, particularly in countries where complex legal structures or ministerial dirigisme can be observed.

While close interaction with the political and budgetary authorities is required, the legislation should specify the parties involved in and procedures for planning statistical programmes. Legislation and the final perimeter of statistical outputs should be left to political decision-makers, but decisions on methods, standards and procedures and on the content and timing of press releases should remain in the hands of the statistical services. However, in one country the statistical institute itself reports, and in six others stakeholders have pointed to, shortcomings in the content and timing of releases, multiannual programming and the role and status of the Director-General.

Strong legislation underpinning the professional independence of statistical services is a necessary condition for good governance, but is not sufficient on its own. For example, a revised Statistical Law is now in place in Greece, but implementation must still be carefully monitored, as it takes time to change the administrative culture. On the other hand, in a few countries history and tradition are considered to induce de facto professional independence, even if the legislation does not fully comply with the Code [of Practice]. This was also assumed to be the case in Canada but proved wrong (7-8).

In an interview by Sharon Broadfoot in the Ottawa Citizen, this is expanded upon.

"We were utterly astonished, given our view of Canadian statistics. We didn't expect it to happen in Canada, quite frankly," said Johnny Akerholm, chair of ESGAB. "We've all been full of admiration of everything that is going on in the statistical field in Canada. Canada has frequently been seen as the benchmark, the best performer."

ESGAB was established by the European Parliament in 2008 to boost the professional independence, integrity and accountability of European statistical agencies. One of the tenets of the organization's code of practice is that the autonomy of statistical agencies should be guaranteed by legislation, and the annual report cites Canada as a country where the statistical agency had a tradition of independence, until the government exercised "dormant legal powers" in making changes to the census.

[. . .]

Greece provides a recent example of the importance of reliable statistics produced free of political interference, Akerholm said, noting that the country's statistics obscured the true depth of financial troubles that are now rippling through the European Union.

"Of course, the figures might be all right even if you have a political influence, but there could always be the suspicion," he said.

Note the linkage of Canada with Greece. This is not a good thing. And yet, as Don Cayo of the Vancouver Sun notes, Canada's situation will look brighter with the tricks of the new voluntary survey in much the same way that the tricks of Greece's government-influenced statistical agency made that country's economic situation look so much better.

We'll be seen to be richer than we were just a few years earlier, not to mention better educated and more universally able to communicate in Canada's two official languages.

Of course, it won't be true. This will be a distorted picture painted by the 2011 census. Thanks to the federal government's decision to scrap the mandatory long-form questionnaire, it can be expected to seriously under-represent huge groups of lower-income Canadians who'd make the picture grittier -- not to mention more realistic. And much more useful.

Of course, even though it's fairly easy to predict which groups are likely to be under-represented, there is -- without the mandatory long-form data -- no way to know the magnitude or distribution of the under-count for any given group. This matters.

"The census is used enormously widely," Fellegi told me when we talked shortly after a private meeting of the cabinet-appointed National Statistics Council last week. "City planners, for example. Or business people who want to decide where to put a plant, and whether they can find the kind of labour they need in a neighbourhood, or where they should open another retail outlet or shopping centre. Or a school, for that matter, or an immigrant assistance centre, or a home for the elderly."

John Richards, a public policy professor at Simon Fraser University and a member of the Statistics Council, noted that StatsCan has gone to unusual lengths to figure out who it's most likely to miss when the long-form survey becomes voluntary. It has reexamined its 2006 data from three cities -- Toronto, Winnipeg and Bathurst, N.B. -- to see which respondents replied readily and which ones had to be chased and cajoled. [. . .] What this study showed first is that it's the smaller centres -- places such as Bathurst, with about 15,000 people -- whose results are most likely to be seriously skewed. And those results will under-count immigrants and aboriginals in particular, as well as some other lower-income groups.


Anonymous said...

"City planners, for example. Or business people who want to decide where to put a plant, and whether they can find the kind of labour they need in a neighbourhood, or where they should open another retail outlet or shopping centre. Or a school, for that matter, or an immigrant assistance centre, or a home for the elderly."

Which of the above doesn't belong? If the census is to forward a legitimate gov't need, so be it. But I (and many others) don't want to compelled by force of law to help provide Starbucks with cheap market data.

Randy McDonald said...

How many others? Was the census even an issue for people before the government raised it?

As for marketing data, public domain use for that data is entirely legitimate, whichever domain is involved.