What to Do When StatsCan Geography Doesn’t Fit
Have you ever ordered or used custom census geography to do analysis? How did the geography fit? If you could do it over again, would you have used a different method for establishing the smallest units of census geometry?
Not long ago, I was asked to speak with a representative from the national census agency of a relatively small, but populous, Asian country about Ontario’s view of Statistics Canada geography. It seems that their census geography aligned with their internal provincial and municipal boundaries, but that the resultant statistics numbers were too large to permit useful analysis. They wanted to develop a completely new low-level census geography that would meet their needs, something like the Canadian Dissemination Areas and Dissemination Blocks. The representative already had meetings booked with StatsCan geography staff in Ottawa to discuss general principles, but she wanted a user perspective as to whether the smallest geographic units of the Canadian census truly met their objectives. Through a series of short meetings with StatsCan Ontario Region and a number of Ontario ministries we learned about working with StatsCan data.
The StatsCan Ontario Region staff, who, by the way, were very helpful, gave us examples of why people would order tabulations of custom geography that did not align with existing dissemination boundaries and explained how they performed the calculations. They also showed us how the suppression and rounding rules to protect privacy results in the sum of statistics for a number of smaller units never equaling the custom tabulation for the same area.
One of the most unusual meetings was with the Ministry of Natural Resources (MNR). Here, it became clear that it will never be possible to develop the perfect low-level census geography to meet everyone’s needs. MNR is frequently asked to calculate the economic impact of preserving a particular rare or endangered species, most commonly by preserving its habitat. Aligning boundaries with the feeding and breeding areas of species is definitely not within StatsCan’s mandate. How do you use StatCan data to answer such questions?
The short answer is that it’s not easy. The process generally starts by trying to determine all primary and secondary economic activities that affect or are affected by the species in question, or by preservation of its habitat. Initial activities include things like tourism or resource extraction. Secondary activities include things like transportation routes that would be curtailed or not built, and chains of events like whole towns losing their mills or mines and no longer able to support local businesses, or even, in extreme cases, reduction in national output of a particular commodity. Then comes the hard part: trying to determine exactly what statistics are affected for what parts of what census geography. Often, in rural and far north areas, StatsCan areas are geographically large so as to include enough population data to meet privacy suppression rules. The consequences of, say limiting logging in one area may result in reduced output of a mill in another area or the overall reduction in lumber being distributed through a third or aggregated area. The large boundaries used by StatsCan in the north may not be sufficient to identify these results.
Saving biodiversity, and in particular, specific species is definitely the expected outcome, but knowing the facts is essential to be able to make data-based decisions and to anticipate the inevitable arguments. Parsing out Statistics Canada data and working with custom geography is one very important part of the process.
We also visited with two different sections of the Ontario Ministry of Health & Long Term Care (MOHLTC). Both sections use StatsCan data and experience different challenges. One section that works with Local Health Integration Networks (LHINs) and the other works with public health units.
For those of you who are not aware, the last couple of years have seen a massive shift in the way the Ontario Government delivers primary health care. Many of the direct responsibilities are now being handled by the LHINs rather than directly by the province. When these LHIN’s were originally set up, the area boundaries were explicitly aligned with StatsCan census geography boundaries. Due to the large amount of statistical analyses expected to be performed on a regular basis, this made perfect sense. Imagine their surprise when, at the very next census, StatsCan moved their boundaries to better represent changes in population density! So, despite the best laid of plans, it meant going back to using custom geography.
With the second health section, the geography had always aligned with the boundaries of local municipalities, which had very rarely, if ever, aligned perfectly with StatsCan boundaries. For the most part, StatCan’s concept of urban areas meets the needs of statisticians rather than municipal authorities. This fulfils the mandate of Statistics Canada; however, this sometimes makes it very difficult for local government to perform analyses without a bit of tweaking. One of the biggest concerns of local government is the current model used to define municipal boundaries. StatsCan only follows roads, while many municipal distinctions are divided by rivers, water bodies, railway lines, utility corridors or property lines. Once again, the need arises for custom geography.
We met organizations that were frustrated that most statistical data was not available at the standard, low levels of geography, making it difficult to target their services to, for example, meet the local language or cultural needs. Some organizations complained that the date of the once-in-five-year census was always in May, when students were no longer attending the local post-secondary institutions. Some users were unhappy that they could not total the statistics for several low-level geography units to get dependable values for do-it-yourself custom geography. However, the vast majority of experienced users of StatsCan Census admitted that there was no perfect method of devising a low-level geography that would ever meet everyone’s needs while protecting the privacy of the population.
Do you have any suggestions? Do you or your organizations have a better idea or practice?
Take care and stay well…
Raphael Sussman
Comments are closed.