Monday 19 July 2010

Long form census

Canada must be a pretty damn boring country if the big issue of the month is scrapping the mandatory long form census. Still, while it might be a dry and esoteric subject, there is some significance to it. The significance is not just in the potential loss of accuracy in the data, or increased costs of collecting the voluntary surveys that will replace it; but in the bigger picture about how people view the government. Anytime the government implements a policy that costs more money, with explanations that don't make any sense at all, people get suspicious and cynical; and start hypothesizing about alterior motives -- very much like the Manitoba government and their unjustifiable direction regarding Bipole III.

Let me paraphrase some of the arguments:

Tony Clement: "We are scrapping the census because Canadians are complaining about the intrusion of their privacy."

Office of the Privacy Commissioner: "Uh, no, actually. We've only received three complaints in the last decade."

Tony Clement: "It was a recommendation from Statistics Canada. If it's good enough for them, why isn't it good enough for you?"

Statistics Canada: "Uh, not quite... You asked us for options, and that was the least recommended of the options we gave you."

Let's let Maxime Bernier get in on this:

The state does not belong in the bedrooms of Canadians and Canadians must have the freedom to choose if they want to answer it.
Trudeau? Really, Maxime, you're going to pull a page out of the Trudeau playbook?

Then there was Dean del Mastro, who "has taken a number of statistics courses" in his life. Dean explains that, in fact, voluntary data is used all the time, and through the magic of random sampling, can be quite accurate within know parameters.

Hey, I've got a riddle for you: what has two thumbs, a green avatar, and has also taken a number of stats courses?


Me. It's me you idiot.

Maybe that would have worked better if you could see me. Anyhow, Dean is quite right of course ... voluntary surveys are used all the time. Money is spent and decisions are made based on the results of those surveys, and there is a whole science concerning the design of questionnaires and surveys, and the accuracy and biases therein. So why can't it be used in place of the long form census?

Well, for one, it's going to be more expensive because it's going to go out to more people. Also, if asking 1 out of every 5 people personal questions is wrong, is asking twice as many people personal questions not wrong?

There is also a question about potential biases: when response rates drop, accuracy becomes more questionable. With the sensitive nature of some of the questions on the census, one might expect response rates to drop significantly. This is more problematic when the response rates vary between different demographic groups. You might be able to adjust for that in some cases, but the long form census itself is one of the key sources for collecting that demographic data, and therefore is very important.

To me, this decision by the government is puzzling and disappointing. They still have failed to properly explain it, choosing instead to make unprovable claims about privacy concerns, and playing up the fear factor about soldiers smashing down your door and throwing you in the gulag if you don't fill it out.

My answer to them is this: if some of the questions on the long form are too personal and have outlived their usefulness, then change those. Shorten the form a little bit. Have a little chat with Stats Can and say, look, is this particular question really necessary? But you've lost this battle, and anything you say at this point will just make you look more foolish. Not only that, but Michael Ignatieff could run over a baby on his bus tour and nobody would notice because of this stupid census squabble.

Or maybe that's the point ... keep the press's attention turned away from Ignatieff's BBQ road trip, even if it means making a bad decision and creating a whole new wave of cynicism in the electorate.


Gustav Nelson said...

Really though, why does the government need to be collecting this data? Can and don't private companies already collect data on their own?
Really then, why are taxpayers paying for work that private companies should be doing?

cherenkov said...

The short answer is no ... private companies don't collect this data. They collect data, but not this comprehensive or consistent, nor is the data collected by private companies available for use by government agencies, universities, etc..

Gustav Nelson said...

Perhaps because government already collects this data, there is no need for private companies to do so.
If government were not to do so, I'm sure there is someone in the marketplace that is willing to collect this data and offer a price for it.

Anonymous said...

@ Gustav: don't be thrown off by the red herring that the govt's supposedly collecting all this extra info primarily for private biz's benefit.

They're doing it primarily for the various levels of govt's benefit for social planning purposes: to track things like unaffordable housing, poverty levels etc: topics they analyze like

Private Biz's aren't going to bother -- and certainly aren't going to pay for -- tracking social topics like are parents spending enuff time with their kids, or too much time in commuting or unpaid caregiving of their old parents or neighbours, & how many are living in crappy rundown housing they're paying so much rent on they're just one paycheque away from bankruptcy.

They also use it to properly 'weight' all the other useful surveys they & other agencies do (to project their very small samples to the pop. as a whole), like the monthly Labour Force Survey.

But, yes, biz's, charities -- & prob'ly most importantly in this context -- rival political parties, all do benefit from some of this long form demographic data on income, occ's, ethnicity, religion, etc. But the gov't needs / can put good use to it & should be collecting it, anyway, & they can do so far better, cheaper, more comprehensively & less intrusively (than unleashing zillions more dinner hour polls). And they don't give this info away at the more useful neighbourhood level: they sell it on a cost-recovery basis for around $100-M a year.

So the Fraser Inst. talking pts on this are simply full of it.

/* Google Tracker Code