A census story I wrote almost two years ago at Postmedia has been making the rounds recently because of comments I attributed to Niels Veldhuis. He tweeted today that he “Never claimed that larger sample solves selection bias. That is simply inaccurate and wrong attribution.” There were no complaints from Veldhuis or the Fraser Institute at the time about the story, but I retrieved my interview notes so I could defend my work.
Here are the notes in full—I bolded the direct quote that supports the disputed paraphrase from my story:
Niels Veldhuis – senior economist with the Fraser Institute
“The reaction thus far has been one that you would expect: the academics, economists, other social scientists have jumped all over it and seeing as they’re heavy users of the data, they’re a vested interest group and obviously want to keep access to this data.”
“I think that Canadians ought to have a very sober second look at the mandatory long-form census and really ask themselves to answer what I consider to be very private questions.”
“If you flip this on its head and say there was no long-form census, what would be the reaction to the Conservative government coming out and asking folks to check off whether they’re white, black, Arab or Asian, where they work, how much time they spend with their kids, whether or not they have difficulty climbing up stairs or bending down, which member of their household pays the bills, whether or not their homes have any missing or loose tiles and those sorts of things.”
“These questions are things that I don’t believe the government has any business or authority forcing Canadians to answer.”
“The academics have to ask themselves on what merit the government should force Canadians to disclose all of these personal things. I certainly understand that social scientists, and I’m one of them, like to play with data, they like to analyze social trends and economic trends, but the reality here is there really is no good basis for collecting this information. It’s a cheap way for academics and social scientists to get information that I believe she be acquired using voluntary means.”
The current census is “by no means a complete analysis” and there are “serious problems with compliance” and there are groups left out or underrepresented
If you look at aboriginals, they’re underrepresented
They still have the short form to adjust biases, though he concedes the long-form will not provide the other info from the long-form
“To plan what? This is what should be worrying average Canadians – this information is used by central planners to plan how to tinker with the lives of Canadians.”
Most of the data used in Canada is collected by private firms: polling, focus groups, marketing, etc.
“There is no reason we have to force Canadians to answer these questions. It can be done voluntarily.”
Host of other surveys, much more powerful like Survey of Labour and Income Dynamics, which tracks Canadians in other ways
All sorts of other surveys they can use on a voluntary basis
“beyond the scope of what the census should be doing” to force people to answer
if this was introduced, people would be upset
“The long-form census is not a census of all Canadians. It is a sample of Canadians. It’s only 20 per cent of Canadians who fill out this form.”
“You can be representative and have a voluntary survey if you just expand the sample size.”
The idea that the bias increases with bigger sample size “would go against” most private sector surveys, including those of the Fraser Institute, which capture all sorts of minority groups
They just did a survey on tax returns and asked Canadians about 40 questions in tax returns: how long it took them, what kind of assistance they had, made sure there was a sample distribution across provinces and that you have a sample that’s representative across different groups
Population statistics are what they use to make sure they get distribution across different provinces
i.e. why would it matter if they had distribution across income or ethnic groups?
The long-form census isn’t the only way we get income data, everything that is listed in the long-form data
“All of this stuff the government already collects. We’re forced to provide Canada Revenue Agency with this data.”
They use long-form census data
“The voluntary manner to collect this data will be reliable.”
Certainly, the response rate won’t be as high but you can get a representative sample through voluntary means
“There are still serious biases in the current census, even if you make it mandatory. There’s all sorts of underreporting and data problems with the current census. One only has to go to Statistics Canada’s website to see the types of errors in the census.”
Coverage errors, non-response errors, sampling errors, etc.
They use long-form data for their school report cards, they don’t adjust school rankings based on SE data but they do provide info on average parental education and average income for each district the schools are in
“It’s something we ought to have a closer look at. Personally, I don’t see the reasons for Canadians being forced to divulge this sort of private information”
they’re likely going to come out with something on the piece


