I’m going to go ahead and assume anyone who’s reading this blog knows about the Canadian government’s recent decision to scrap the mandatory long-form national census, and attempt to replace it with a voluntary “National Household Survey” (NHS).
I’m going to assume you all know that the information from long form census is used for all kinds of governmental and non-governmental planning, including social programs, financial allotments for various uses, and, yes, library service planning.
I’m going to assume that readers all have the basic statistical proficiency (that the leaders of the Conservative party apparently lack) to know that a mandatory survey with near 100% compliance given to 20% of the population will almost certainly have greater validity that a voluntary survey of 33% of the population, because even if the absolute numbers of responses are the same under both surveys the response rate in the latter will almost certainly be lower and thus reflect self-selection/non-response bias. This, by the way, is a neat way to create policy-based evidence.
I’m also going to assume you know that the Tories are holding fast to their position, even in the face of the Chief Statistician of Canada’s resignation and overwhelming outcry from people who are not always allies on the issues.
Throughout this whole kerfuffle, the Harper government has insisted that the change is due to privacy concerns. They insist that the long form is too invasive. That people think the state should not be requiring that individuals report such personal information as the number of bedrooms in their dwellings and how we travel to work. Etc.
You probably also know that little evidence has been produced to back up the claim that many people are concerned about the invasiveness of the questions on the long-form (although in the most recent 2006 census there were refuseniks on the basis of the gov’t subcontracting census work to Lockheed, an issue that is going unmentioned today).
Privacy, my foot! This change has nothing to do with privacy. Scrapping the long form does incredibly little to improve privacy of Canadians, and in fact may even make our personal data less private and secure.
Here’s some info about the long-form census, and voluntary StatsCan surveys, in terms of privacy. All of this is written to the best of my knowledge, so if anyone works with census data or has StatsCan connections and can clarify or expand on any of the below points, please do let me know.:
Public Release of Data
Full census data is normally released to the public after 92 years. Since the passage of S-18 in 2005, there has been an opt-in check box on the census, which must be checked in order to release that data after 92 years. Opt-in, while frustrating to researchers and genealogists who wanted opt-out instead, is a pretty high standard of privacy protection. Voluntary surveys are not released to the public, ever. Thus, it would seem that the NHS would be more private, 93 years from now, than the census. Or, rather, it would seem so if Tony Clement hadn’t assured us that they were changing things so the NHS will also be available after 92 years No word as of yet on whether the NHS will have an opt-in box.
Commercial Interest Intervention
Commercial firms do not have access to the census planning process. Commercial firms as well as non-profit researchers may purchase modules for many voluntary surveys from Stats Can (e.g., the Canadian Community Health Survey), however. Should the long form census remain a voluntary survey, I would not be surprised at all if business were eventually granted access to this survey too. In fact, I would expect it, if the survey loses the “sacredness” of the National Census.
While we might never be fully confident that stewards of any data could never possibly misuse it, the census is subject to privacy safeguards above and beyond other surveys. Disclosure for any purpose prior to the 92-year blackout period would be subject to fines and penalties under the Statistics Act, which requires StatsCan employees working with this data to be sworn to secrecy. Unlike other StatsCan surveys, the Census is not available to researchers outside StatsCan as a full microdata file. It is also not eligible to be linked with other databases, unlike other data sources including other StatsCan surveys. Again, should the long-form census turn into and assume the norms of a voluntary survey, these safeguards will likely be lost.
The Short Form & Privacy
The still-mandatory short form, which nobody appears to be speaking out against, provides plenty of information on which to base discrimination (name, age, sex, marital status including whether partner is same or opposite sex, first language learned). Sexual orientation is pretty personal and invasive to ask everyone to disclose on a mandatory basis, if you ask me — perhaps more so than many of the questions on the long-form. First language-learned gives a lot of ethnicity & immigration information, even without the long form asking about where your ancestors came from.
As comment poster LindL on the Worthwhile Canadian Initiative blog writes, of the move to scrap the long form but retain the short form,
“If it’s wrong, it’s wrong. What you’re arguing is equivalent to ‘Stealing is wrong. So I’m not going to steal a car, instead I’ll just steal a bicycle’.”
Harper’s Privacy = Girl-Power Barbie
The Harper government is using privacy as a red herring here. Calling these changes to the census good for privacy is like calling a “girl power Barbie” feminist – in other words disingenuous cooptation of a real issue in order to improve market share. I fear that to take the Harper/Clement argument at face value – that this change is about privacy – is to play right into their hands.
What is the Harper government’s interpretation of Census privacy? Well, apparently that applies to the government need/ability to make private and unseen the concerns of Canadians regarding the census changes! The long-form census, although second top rated issue by participants in the recent digital economy econsultation, was buried on the site. (Although I wouldn’t vote for this very specific issue to be the top concern of Canada’s digital economy, I am quite impressed that it got so many votes, continuing to add votes and hold second place *even after it was buried* on the site and could only be accessed via a direct URL.)
As a privacy and social justice advocate, the long-form census is not what I am worried about. There are a lot of less secure sources of data out there, with less redeeming social value to worry about. For me, the benefits of responsible, privacy-sensitive data collection and stewardship sometimes outweigh the risks, but I can respect opposing opinions.
It’s one thing if you oppose any government collection of personal data. That I can respect. I think it’s a perspective that usually comes from a position of privilege, and I don’t take this stance myself, but I can respect it if you think the long form should be scrapped along with a whole bunch of other things, on principle. But THIS? Only scrapping the long form? That is not about privacy. If you want to campaign for the end of all government information gathering, fine, but the census is really not the place to start. And don’t think for a moment that the Conservative government is on your side an end to government information-gathering
However, in a practical sense, this change is not improving the privacy of Canadians in any measurable way, and is in fact eroding the privacy of those who answer the voluntary survey as well as hurting those who don’t by virtue of providing a skewed and unreliable demographic profile of Canada.
On the upside, if they do go ahead with this plan to convert the long-form to voluntary National Household Survey, I think it’s likely that whoever is in power in 2016 will change it back. In that case, while we’ll lose the 2011 data for planning and research purposes, it will be interesting to see how and to what extent making a survey voluntary creates deviations from trends (i.e., we may be able to tease out which subgroups will & won’t respond to a voluntary survey). A natural experiment in the making!
ps – For whatever it’s worth, I agree with Sandra Finley that the census software etc should not be subcontracted to Lockheed Martin – much like I do not think the BC MSP data should be with Maximus. But that’s another kettle of fish.
pps- Tracey over at datalibre.ca has been researching the census as part of her PhD research and has been chronicling this recent debacle quite comprehensively.