The vast majority of Canadians were furious when, in the summer of 2010, the federal government of the day decided to exclude the long-form questionnaire from the 2011 census. They were enthusiastic to fill out their long-form census questionnaires. They had practically no security or privacy concerns.
At least that was the popular media narrative.
Only a few months later, Canadians handed the same government that made that seemingly fateful decision its first majority. Despite their supposed enthusiasm, one in three Canadian households opted not to complete the voluntary 2011 National Household Survey (68.6% unweighted response rate) – and that’s with Statscan spending tens of millions more on ‘follow-up’ and accepting forms with as few as 10 of 84 questions completed. As it turned out, security and privacy were the primary reasons prosecuted census refuseniks offered for their refusal to comply.
Canadians apparently not so enthusiastic about census
Five years on, the census has been ‘restored’, and the same popular narrative persists.
Despite supposedly being more enthusiastic than ever about filling them out, Statscan has indicated less than half of Canadian households had submitted their 2016 census forms by the May 10 deadline; the long-form response rate is likely significantly lower, but the agency hasn’t had time to ‘research’ it apparently.
Google has indicated that the second most popular 2016 census-related web search result on May 3rd – the day after Statscan’s census mailing – was “What happens if you don’t fill out the Census in Canada?”; the fifth most popular was “Is the Canadian Census mandatory?”
Even more concerning, a significant share of the top comments submitted in response to a CBC news census item on May 2nd were from individuals indicating they would either refuse to complete or otherwise spoil their census questionnaires if compelled to complete them; one top comment read:
If I am forced to provide my personal information to the government, I will do my best to provide the most misleading and inaccurate information that I possibly can.
Significant change to income question, informed consent
While the media was quick to jump on the 2011 long-form census cancellation announcement-by-omission in the June 2010 Canada Gazette, the February 2016 Canada Gazette item announcing a significant change to the 2016 long-form received little scrutiny.
The 2006 long-form census was the first to include an option for Canadians to forego responding to questions on income by giving Statscan consent to access their Canada Revenue Agency tax return data; this was repeated on the 2011 National Household Survey.
A brief note in the Canada Gazette, dated February 6, 2016, announced that for the first time ever Statscan would neither be asking Canadians to fill in income questions nor for consent to access their CRA tax return data; instead, it would proceed to link their CRA data to their census forms without asking. Further, this change would apply to all Canadian households, both those receiving the short as well as the long form questionnaire.
While Statscan will be quick to point out that most long-form respondents over the last two census cycles had given their consent anyways – 82.4 percent in 2006 and 73.2 percent in 2011 – that’s little reassurance. Unfortunately, most Canadians don’t take the time to research and understand the implications of volunteering their personal information.
When asked about Canadians’ willingness to give the agency consent to access their tax records, a Statscan income analyst once confided, “If they knew how much information they’d be giving us access to, they’d probably think twice about it.” (Notably, that analyst demurred when asked for specifics and whether he would personally give consent.)
It’s likely for this reason Statscan never clarified on the long-form questionnaires in either 2006 or 2011 the scope of information it would collect from consenting respondents’ CRA tax records. (And likely why requests for a copy of the information sharing agreement between Statscan and CRA go unanswered.)
Citizen profiles using 2016 census and social insurance numbers
However, a quick search of Statscan’s recent pages turns up a note under Terms and Conditions > Privacy Notice > Record Linkage > Approved Record Linkages > 2014 announcing its use of SIN numbers to create a Social Data Linkage Environment:
The DRD (derived record repository) is created by linking various Statistics Canada data files for the purpose of producing a list of unique individuals…
Some of the data files used for the DRD include the Census of Population and National Household Survey, T1 Personal Master Files (Tax), Canadian Child Tax Benefits files, Vital Statistics – Birth Database, Vital Statistics – Death Database, the Landed Immigrant File and the Indian Registry…
The DRD would initially be comprised of the following personal identifiers: Surnames; Given names; Date of birth; Sex; Marital status; Date of landing/immigration; Date of emigration; Date of death; Social Insurance Numbers (SIN), Temporary Taxation Numbers (TTN), Dependant Identification Numbers (DIN); Spouse’s SIN/TTN; Dependant/Disabled individual SIN/TTN/DIN; Parent SIN/TTN; Health Information Numbers; Addresses; Address Registry Unique Identifier; Standard Geography Classification codes; Telephone numbers; Spouses’ surname; Mother’s surname; Father’s surname; Alternate surname and a Statistics Canada generated sequential identification number for each individual identified through the annual DRD linkage process.
So Statistics Canada is in fact creating a “citizen profile in all but name”, and doing so with SIN numbers, ignoring past advice from the Office of the Privacy Commissioner.
Statistics Canada response
Many Canadians – likely in the 20 to 30 percent range, the same ratio that did not consent to Statscan’s CRA linkage requests in 2006 and 2011 – upon receiving their census form this time around would justifiably have had second thoughts upon reading the note from the Chief Statistician stating that their tax records would be accessed without their consent.
The February 6 Canada Gazette notice included the contact information for Marc Hamel, Statscan Director General of Census Management.
When asked about whether the Office of the Privacy Commissioner had been consulted prior to the change in consent to access Canadians’ CRA tax record data, Mr. Hamel responded:
“The Privacy Commissioner is always kept informed of all the plans for census of population. Because of the magnitude and importance of the exercise, we have regular conversations with the people at the Office of the Privacy Commissioner. So all of the changes to the planning and procedures for the program were discussed with the Privacy Commissioner. On our part, either individual Privacy Impact Assessments [were] conducted for the census or integrated into the Statistics Canada Privacy Impact Assessments for our survey work.”
Asked why the change was made on what appeared to be rather short notice – less than three months prior to census mailing – Mr. Hamel noted:
“The plan for the income was actually always to proceed this way. So this has been going on for a couple of years at least, it’s been tested in various forms; it was part of our testing in 2014. We also tested communication, like how the messaging to the population takes place. The approach has been used on some of our household surveys in the past, very successfully. And it’s also based in reaction to past experience with the census itself, where the vast majority of Canadians were granting us permission to do this before, and also public reactions to Government of Canada asking for information it already has in its possession, basically.
When asked why there was no mention of privacy consultation regarding the referenced change in the 2016 census consultation reports, Mr. Hamel indicated he didn’t know the answer off the top of his head, but offered to follow up by e-mail.
Since the census hotline was so busy on May 13th – three days after the census deadline – that callers had been receiving messages all day directing them to try calling back later, Mr. Hamel was asked whether many Canadians had called with questions or concerns regarding the referenced change.
“I haven’t had any reports that this is a common question. The line has been very busy, so we have been receiving a very high volume of calls – primarily people requesting assistance in completing the forms, so… but I haven’t had any reports that we’ve had a lot of questions in relation to this aspect.” (emphasis added)
As to the question of why Statscan had sought informed consent to access Canadians’ CRA tax record data in 2006 and 2011 but did away with asking this time around, Mr. Hamel indicated he would respond to the question after receiving it in writing. To date, no response has been received.
Statistics Canada response in context
A few issues with Statscan’s response. First, it would be impossible to tell how Canadians would react to being notified that Statscan would be accessing their tax records without their consent, and that they could face fine and/or prison if they disagreed with the arrangement. The agency could just assume a response is a response; however, as the previously referenced CBC comment suggests, coercion has an unknown and decidedly negative impact on response quality.
The 2016 census program content test report, released on April 1st (seriously), indicates the 2014 census short form test questionnaire return rates ranged from 81.2 to 88.5 percent, while the long-form return rates ranged from 68.0 to 80.2 percent. Remarkable at first glance, since less than half of those surveyed responded to the 2008 long-form census test questionnaires. All survey tests, as well as all household surveys except for the census, are voluntary. Or at least they’re supposed to be.
The remarkable 2014 return rates apparently came about as a result of Statscan (mis)representing the census test as mandatory. While obviously questionable from an ethical standpoint, it gives a good indication of what the expected response rates for the actual census will be. And that’s not good.
To Mr. Hamel’s specific response that the change to the income question and consent was tested, the 2016 census test report explicitly states:
Income data were not part of the 2014 Content Test. The forms that were tested did not contain a section on income, neither in the paper version nor in the electronic version.
To Mr. Hamel’s assertion that this change was implemented after submission of one or more PIAs and consultation with the OPC, the latest (2014-2015) annual review by the OPC includes only one reference to the census – to Statscan’s questionable Social Insurance Number question on the 2014 census test, which the agency had not consulted the OPC on in advance. The Privacy Commissioner reiterated his office’s negative view on SIN collection, and noted Statscan’s assurance that it would not proceed to ask respondents for their SIN as part of the 2016 census. It appears the OPC was unaware of Statscan’s plan to take it without asking.
Privacy Commissioner response
When asked to comment on any consultations sought by Statscan on changes to the 2016 census, an officer from the OPC explained the conspicuous absence of information in his office’s annual review:
“I spoke with my colleagues about this not too long ago because we had been getting some requests, some calls, about the census. To my knowledge, from everything that I’ve looked at, no, we haven’t had any direct consultations about the 2016 census.”
When asked to elaborate on the nature of the calls his office has been receiving, the OPC officer remarked (without prompting):
“Some individuals had concerns with the information that was being collected from Canada Revenue… My understanding is that in previous censuses that was collected from individuals, and there was a change made by an Order in Council for the 2016 census, and the decision was for Stats Canada to collect that administrative data from Canada Revenue.”
When asked whether questions and concerns over Statscan’s collection of CRA tax record data without consent comprised a ‘significant share’ of 2016 census-related inquiries received by the OPC, the officer confirmed, “Yes, yes. I think it’s fair to say it is. It’s certainly a decent portion of them.”
When asked for an opinion on the change from asking for consent – as Statscan did in 2006 and 2011 – to simply taking people’s personal information without asking this year, the OPC officer spoke generally about the Privacy Act:
“Consent in the Privacy Act is an underpinning element as opposed to being in a sense a more structured idea. If someone were to look at section 5 of the Privacy Act, what 5.1 says is that a government institution shall wherever possible collect personal information that is intended to be used for an administrative purpose directly from the individual to whom it relates, except where the individual authorizes otherwise, or where it may be disclosed under subsection 2.”
The OPC officer went on to add, “I’m only giving the Privacy Act. There’s also the Statistics Act and how that informs what they do. And also any information sharing agreements that they potentially have with other departments…”
Many Canadians justifiably concerned
As discussed earlier, Statscan refuses to publicly disclose the content of its information sharing agreement with CRA. As discussed previously, Statscan also does not consider what it does – administering surveys – an ‘administrative act’, which would effectively exempt it from the Privacy Act altogether (and also from the government rules regarding the collection and use of social insurance numbers).
Between the seemingly contradictory information provided by Statscan in response to questions regarding privacy and its decision to suddenly eschew informed consent when collecting people’s personal information – without either public consultation or consultation with the Privacy Commissioner – Canadians should justifiably be concerned.