Monday, 26 July 2010

From the Mailbag

Regarding the census, Ken Breadner commented:

I just broached this with my wife, who used to run a successful market research company. I was stunned to find she sides with Harper and Clement on this. Her argument is that quite a lot of census data is essentially useless a week after the thing's taken. Especially economic/employment data, in this volatile economy. She also independently brought up the falsification of data, which she believes is very high due to (a) people actually resenting the fact it's mandatory and (b) a number of respondents who do not understand the questions. I asked her whether this falsification was equal to or greater than the degradation you would get from a voluntary census, and she said "quite possibly". Now, I'm not going to put my wife up against the thousands upon thousands of highly educated people who disagree vehemently with her. But I have to admit I was surprised to find support for this that didn't come from a stupid person, and right in my own house.
Now, Eva is not sure what to replace a census with. But she says printed surveys are among the worst tools one can possibly use if accuracy is any concern at all.

What may be inaccurate data for one person may be perfect for another. It all depends on the data being asked and what it is being used for. If you need time dependent analysis, then a 4 year old census is not it. From what I understand, polling forms that are constantly querying for time dependent information use the census (when they can) to calculate their degree of error. Because polls are voluntary, they need a known baseline (the census) to correct for error in their voluntary results. If the census were voluntary, we would have no baseline to calculate error. All results would in effect would be some group of X number of people agreed to Y. How that translates to a percentage of the population we have no way of knowing. The only percentage of the population that agree to Y are the people that answered the question.

If I've screwed up the above explanation, please clarify in the comments.

It still is a fact that all of the experts agree that a voluntary census is less accurate. The US found this out after doing a study to see what the effects would be. (Imagine that! Using data to verify a proposed policy. Will wonders never cease.)

So if our census has accuracy issues, we need to find a way to make it more accurate. Or find other ways to correct the data after the fact (such as regular polling). The European model that many conservative supporters are pointing to (there is an irony there) use a cradle to grave registry that tracks all citizens all the time. No census needed because big brother is watching you anyway. That provides the most accurate data of all.

Now would Canadians be willing to accept that solution? I think even the conservative mouth puppets supporting this latest fiasco would rise-up to block that. Unless they truly are incapable of independent thought from the party line.

PS: Eva is not a conservative party mouth puppet. Trust me. Her IQ vastly exceeds that of the mouth puppets that were populating the Macleans message boards. Ken is not inclined towards violence, but he would (deservedly and with full justification) hunt me down and violently correct me if I implied otherwise.


Ken Breadner said...

"The only thing that really bothers me", Eva says to me, "is that nobody ever asks me what MY opinion is."
"Huh?" I blurted, and then went "oh, yeah, just because you said something with what seemed like utter conviction, doesn't mean jack squat".
She smiled. "What I BELIEVE", she said, as if I had remained silent, "is that the census SHOULD be mandatory. We've been using the bloody thing so long now that we couldn't do without it. And the data should be as accurate as possible." Market research companies, she told me, don't tend to bother with the census very often, on account of it rarely intersecting whatever they're studying. When they do use census data, they are acutely aware that it's (a) likely out of date and (b) obtained, on some level, by coercion. Which renders the data at least a little suspect all by itself.
Polling, yes, the census is used when it can be exactly as you describe. There is a surprisingly complicated science behind margins of error.
And no, Eva's not a Conservative mouth puppet. (Thanks for that.) Or a Liberal one, for that matter. I'm the political entity in this house; she only pays attention when I start steaming.

Catelli said...


Thanks for the clarification. Apparently 1 degree of separation from Eva to the comment was one too many!