Changes distort the census

Posted on July 9, 2010 in Governance Debates

Source: — Authors: – Technology

July 9, 2010.   By Dan Gardner
Admittedly, the census is not the sexiest topic, but it is important. The data generated by the census are the foundation of almost every public policy. Social science in this country would come to a shuddering halt without those numbers. So would a great deal of business. Anyone interested in reality — and I hope that includes every politician and citizen — is indebted to Statistics Canada and its bean counters.

Anyone interested in reality should also be disgusted with the Conservative government’s latest act of gratuitous stupidity.

In the past, the census has consisted of a short survey which goes to every Canadian household. Its questions were basic — age, sex, number of people in household — and everyone was required by law to fill it out and return it. A second survey was longer and asked questions about everything from education and income to ethnicity. It was sent only to one-fifth of households but it, too, was mandatory.

In the forthcoming census, the short survey will still be sent to every household and it will still be mandatory. But the government has scrapped the long, mandatory survey. Instead, there will be a long, ‘voluntary’ survey. To make up for the expected decline in responses, the long survey will be sent to more households than in the past.

Like I said, this isn’t the sexiest topic. But it is important. Bear with me.

The critical problem in any survey is “sample bias.” Stand on the sidewalk outside a Liberal convention and ask the people how they feel about Stephen Harper and you won’t get an accurate picture of how Canadians feel about Stephen Harper because the people you ask aren’t representative of all Canadians. Your sample will be biased, in other words. Conduct the same survey outside a Conservative convention and the sample will still be biased, although it will tilt in a different direction. Do it at a hockey game — or church picnic, school assembly, or whatever — and the bias may be less obvious. But it will still be biased in all sorts of ways.

One solution, obviously, is to spend a huge amount of money and survey all 34 million Canadians. Another solution is to contact a small sample of Canadians selected at random: Done properly, random selection produces a sample that reflects the group as a whole.

In effect, the census uses both solutions, by sending the short survey to all households and the long survey to a randomly selected sample of households. But still there’s a problem.

If people are free to answer the census or not, some will and some won’t. If everyone were just as likely to answer or not — say, if we all decided by flipping a coin — that wouldn’t cause any trouble because you’d still get a representative sample. But people don’t decide by flipping coins. And so bias will creep back in: People of a certain education level may be more likely to respond than others, or people of a certain age, people of a certain ethnicity, whatever. The bias may not be obvious but it will be there.

That’s why the census is mandatory. It has to be to make the results as accurate as humanly possible.

By making the long survey voluntary, the government has ensured that the accuracy of much of the data generated by the census will decline. Increasing the number of households that receive the survey won’t fix that in the slightest. It’s like saying, when you are standing outside that Liberal convention asking people how they feel about Stephen Harper, “golly, my survey is biased. How can I fix that? Hey, I know! I’ll ask more people coming out of the Liberal convention with big red buttons on their lapels!”

“There is no exaggerating the boneheadedness of this decision,” wrote economist Stephen Gordon, who brought the change to light on his blog. It’s as if the sampling design had been “outsourced to drunken monkeys.”

Apparently, this sort of talk has annoyed the drunken monkeys in question. When journalist David Akin mentioned the controversy on Twitter, Industry Minister Tony Clement tweeted the following response: “Here’s hoping you’ll also cover the other side of census issue, like the coercion and intrusiveness aspects.”

Many horrible puns involving the word “twit” occurred to me when I read that. But I shall resist. Because, as I said, this is important stuff. And the twit’s tweet — sorry, sorry — is the only rationale on offer: Apparently, the long mandatory survey was scrapped because it offends the staunch libertarian principles of the Harper government.

Yes, the staunch libertarian principles of the government. The Harper government. The government that thinks marijuana decriminalization is a Marxist plot, an adult who agrees to consensual sex in exchange for money should be imprisoned, the police did a fine job at the G20, and Omar Khadr can rot in a tropical gulag.

But requiring citizens to fill out a form which is absolutely essential to sound public policy and social science? An outrageous violation of individual liberty.

Stephen Gordon was too kind.

< >

Tags: ,

This entry was posted on Friday, July 9th, 2010 at 12:20 pm and is filed under Governance Debates. You can follow any responses to this entry through the RSS 2.0 feed. You can skip to the end and leave a response. Pinging is currently not allowed.

Leave a Reply