Statistics Canada suspends credit searches, delays plan to obtain banking records | General financial discussion | Discussion forum

Blue said
I got very good detailed replies from my liberal MP. So good experience there. He actually agreed with my concern. And said he was going to move all concerns forward, and pointed out additional people to contact if it becomes necessary. Since Statistics Canada can easily find a fully anonymized way forward to do their analysis, then hopefully this issue doesn't come up again.  

It isn't as easy as that. The StatsCan analysts can (and would) use an anonymized dataset. But StatsCan still needs identifiable data (probably by SIN, with other crosschecks) from the individual institutions to put together that anonymized dataset.

Again, it is important to remember that neither the original data nor the anonymized dataset would be available outside of Statistics Canada; the original data would be worked on by a very limited group of people (and I believe the original identifiers might be removed after the dataset is created), and even the dataset would only be available to the appropriate groups of analysts. It isn't some sort of "government file" sitting around for bureaucrats to look at. Neither the RCMP, Justin Trudeau, nor anyone else can compel them to release the information they may have for, say, Mr. Tin Foylehat.

Unlike records in bank or other institutional settings, this information never needs to be retrieved through any sort of public interface, so there is much less threat surface for attacks, which would need to somehow access into the secure internal systems. Anything is possible, but I'd be much more worried about my data leaking out from, say, banking systems. Also, while they might (although perhaps not) want to know that the information from Ms. SIN 123456789 for expenditures from Bank#1 is a different credit card than from Bank#2, the card# itself likely could be suppressed or tokenized at the bank's end, and not need to be transmitted to StatsCan.

There are probably things they can do to reduce the sampling footprint, or the range of data, and certainly better communicate their goals and intentions. But the data products produced by programs like this are very important for public policy development and evaluation. Poverty measures, CPI baskets and weightings, design and measuring of social programs or financial initiatives, and many other things greatly benefit from good statistical measures.

StatsCan already collects data through things like the Household Expenditure or Financial Security surveys, as well as access to tax records etc, although those have problems of non-response and accuracy. Hopefully a full public discussion can result in support for a reasonable program (in whatever form that eventually takes) to enable important statistical products to be produced.

Top It Up said
How about them knowing how often you visit strip clubs, buy booze, buy weed, visit a shrink, make cash withdrawals at casinos ... you OK with them knowing that information?  

Well, I'd have a certain amount of concern with employees at banks or credit agencies, with systems that are designed to bring this information up on their screens, wandering through my info for giggles, or passing it along to their friend the cop. A handful of sworn StatsCan professionals during the preparation of the dataset, who don't give a damn and aren't digging in any case? No, I'm not worried.

Again, there are (other) perfectly sensible reasons to question and perhaps oppose this in principal or the details.

Of course, everyone is entitled to their own opinions as to the motives and capacities of institutions (or even individuals). I think most people will understand that the Census still involves a large physical collection effort, and while the temporary employees are trained and sworn, there are inevitably going to be a few (obviously embarrassing) glitches in the field. The commentary tends to take them to task for their lack of transparency in reporting these. I also notice that the articles referenced say things like "StatsCan has a strong track record on ethics and security safeguards, "This is not to say the agency is habitually negligent – it’s not." and "We have always been at war with EastAsia"... oops.

If one is actually interested in the risks of StatsCan electronic data processes, it should be easy to distinguish those from Census field operations and physical records, which don't apply. FWIW, I'm not aware of breaches of StatsCan internal data systems.

The original data with SINs and names and whatnot would be unavailable to all but a small group creating the anonymized dataset, in the secure internal network. It would not be put on jump drives, laptops, paper napkins or anything else that leaves the building. The dataset itself would itself be subject to restrictions, and it is likely that what many analysts would work with is something aggregated in one way or another anyway, depending on what sort of product they are producing.