Have a read of the insightful G&M article “Census decision a slow-motion train wreck“, here is an excerpt,
“The census story is a train wreck in slow motion; the latest car to pile on the flaming ruins is the recent report that Statistics Canada has resigned itself to accepting incomplete responses to the National Household Survey (NHS). [Kempton’s note: Sadly, this first tran]
Many readers may have thought that the census issue was settled last summer; it wasn’t. We haven’t even begun to deal with the consequences of the decision to replace the mandatory long-form census with the voluntary NHS. As Economy Lab contributor Kevin Milligan and his UBC colleague David Green note in Canadian Public Policy, one of the most striking features of the census is its ‘hidden ubiquity’. [Kempton’s note: Milligan & Green’s research note is highly recommended reading. Download the free note and read.] The census is an invisible — and yet essential — element of virtually all the data that inform policy debates.”
An excerpt from “Statscan settles for incomplete long-form surveys in 2011 census“,
“One census enumerator, who spoke to The Canadian Press on condition of anonymity, said workers had been instructed to accept the long forms with as few as 10 of 84 questions answered. They can also declare somebody has given them a “total refusal” simply by speaking to them on the phone.
“We can try and convince them and talk about how it’s a good thing, but a lot of people shut down the conversation quickly when they find out it’s not mandatory.”
Don McLeish, president of the Statistical Society of Canada, said partial responses could cause problems in using the data because some of the analysis is done by looking at the answers to different questions together.
“Certainly some of the useful information obtainable from the long form were things like the relationship among several variables – age, income, housing is one obvious thing,” McLeish said.
“That determines things like whether or not you should put up a seniors’ home in a particular place. It’s the relationship among variables that’s critical. If one of those three things is missing, then that jeopardizes any inference you might want to draw.”
Mr. McLeish adds that if certain questions are left blank for large portions of the population, this could further erode any comparability with previous, mandatory censuses.
Other statisticians and users of census data have warned that a voluntary survey would skew the results because certain segments of the population – the rich, the poor and aboriginals for example – would be less inclined to fill it out.”