Immigrant counts are off – by quite a bit

So far my favourite snippet from the reference material on the National Household Survey (pointed out by this Toronto Star article) is this, one, taken from here:

It is impossible to definitively determine how much the NHS may be affected by non-response bias. However, based on information from other data sources, evidence of non-response bias does exist for certain populations and for certain geographic areas…
(B)ased on the estimates and trends from the sources mentioned above, evidence suggests that the NHS estimate for the population born in the Philippines is overestimated at the national level. According to population estimates, the number of immigrants from the Philippines who entered Canada from January 2006 until June 2011 is 141,502, while the NHS estimate of the population born in the Philippines who immigrated between January 2006 and the survey date, May 10, 2011 is larger (152,270). As well, the population born in Pakistan is suggested to be underestimated… 

Also, that list of townships for which no data will be released? Includes 1814 places. That’s… quite a few.

The Star article quotes Industry Minister Christian Paradis saying the following:

“This was the first time a voluntary National Household Survey was undertaken,” Paradis said. “Our government will be looking at options to improve the quality and reliability of the data generated by the 2016 census cycle.”

Anyone have any advice for him?


National Household Survey and data supression

Statistics Canada has started releasing data from the National Household Survey (the infamous Census long form replacement), along with some information about what they won’t be releasing. This release includes data on Aboriginal Peoples and on Immigration and Ethnocultural Diversity.

Data was suppressed for areas with less than a 50% response rate. No data at the neighbourhood (census tract and census dissemination area) level has been released yet, but there is a list of Census subdivisions (towns) that are being suppressed.  Some of these have populations in excess of 2500, about the size of a census tract (and much larger than a dissemination area), making me wonder how many holes will be in the neighbourhood data and even if they will release it at all.

Here’s the list of suppressed towns:

Update: a researcher tells me that contacts at Statistics Canada have told him that dissemination area data will be available “at least as a custom tabulation.” Hmm.

Update #2: Tracey Lauriault at has a nice summation of news stories on and reactions to the NHS release, including some very cogent criticisms.

Well, we all knew it could be bad. Now we’re seeing how bad.

CCSD Court Case: Equal Right to be Counted

The Canadian Council on Social Development is continuing the fight to save the Census. From their web site:

The fight to save the Census continues as CCSD et al vs The Government of Canada will be heard in the Federal Court on November 23, 2011 at 09:30AM.  CCSD and 12 other partners are fighting for Canada’s equal right to be counted in the Mandatory Short Form, the only mandatory tool left in the group of census surveys that reaches every Canadian.

Visit the site for information on how to get involved.


IASSIST day one – an administrative census?

The IASSIST conference has started – this is the largest annual gathering of data librarians and associated geeks from around the world. The opening plenary was given by Ian McKinnon, chair of the Canadian National Statistics Council, and it was on our favourite topic.

He spoke cautiously. One of his main points was that this government does not back down. His focus early in the campaign against the census was on finding a way for the decision to be reversed that would allow the government to save face – a compromise of some sort. It is too late for that, which means that it is too late to expect a return to the former census status quo under this government.

My own thought is that possibly it was too late from quite early on. Under an adversarial political system, the focus of the opposition parties is on causing the government to lose face, as much, as eary and as as often as possible.

So if there won’t be a return to the status quo under this government, can anything be done?

What Mckinnon suggested was that we see the census change as an opportunity. Statistics Canada has always been reluctant to experiment with changes to the methodology of the census because that would lead to a loss of comparability. But now we have an enforced discontinuity. So why not look at alternatives – take a real look at how best to collect data that meets our national needs for planning and research?

The alternative that McKinnon mentioned is to use administrative data. This is the approach used by some European countries (the Dutch model was talked about in a later session yesterday afternoon). This approach requires a national sampling frame with a unique identifier for each person in Canada to which administrative data could be linked – tax, immigration, employment, etc. Of course, I’m not sure a government with a libertarian bent would really go for such a national tracking system, a point McKinnon alluded to as well. An audience member mentioned the American Community Survey model – a rolling, large sample mandatory survey.

McKinnon ended by emphasizing that the government doesn’t care what statisticians, librarians and other academics think, and neither, in general, does the general public. So our role should be to encourage end users to speak up: health planners, business owners, transportation engineers and others.

Many such groups have already spoken up. So my questions are: can we get them to do more? And can we get anyone to hear?

Census / NHS update

I’m at the National Training Day for the Data Liberation Initiative, which is a Statistics Canada program and part of a week of data-related conferences and meetings taking place. (We also had a CAPDU meeting which I’ll post about separately.) This morning was a session on the Census and NHS, which doled out some interesting tidbits.

Very preliminary numbers on web completion rates of the NHS by people who filled out their Census form online are at about 60%. The Statistics Canada people were positive about this, but IMHO 60% is not good, especially for something that’s replacing a census. They will be doing followup with the non-responders, which may increase responses somewhat. We were told that the 40%-odd non-completion rate includes a small percentage who began their NHS but stopped before finishing. It will be interesting to see how NHS completion by people who got the paper census; my guess is they will be lower.

Groups exhibiting or likely to exhibit non-response bias were discussed; they include people with lower SES, immigrants and aboriginals as well as an additional group I hadn’t thought of before: males under 24.

Apparently the language questions that were added at the last minute to the short census to ward off a lawsuit caused some issues: they had to add another fold to the form to get it to mail, and there’s some concern that recipients won’t unfold fully and notice the additional questions.

Due to lack of resources, there will be no microdata file from the census, only the NHS one will be produced, though it will of course include the census questions. Vince Gray noted that this will make it impossible for researchers to check sample bias in their NHS analyses against a mandatory sample; the concern was noted.

We had also an interesting little discussion on Statistics Canada’s lack of a preservation policy, sparked by some comments from Laine Ruus on the loss of the print census publications.

The IASSIST international data conference will be going on later this week. I’m very much looking forward to the opening plenary on Wednesday: Ian Mckinnon, Chair of the Canadian National Statistics Council, will be discussing the Census issue. The program notes that the “National Statistics Council (is) the group that is tasked to provide Statistics Canada with its best advice. This advice did NOT include cancelling the long form census and substituting a voluntary National Household Survey.”