A Limerick (and a reply to a response to a critique of a study)

John Rose posted a response to my critique of his study this morning. Almost immediately after I was alerted to the posting (via Kerry Gold), we met up to chat at Sweet Obsession cafe. In appreciation of this turn of events, I offer this limerick:

John Rose is a very nice guy

Despite our dispute o’er supply

We just met for tea

And we mostly agree

Where not, please see my reply

He really is a nice guy. And in order to insure our back-and-forth doesn’t become too tiresome, I’ll offer just a quick reply.

  1. Though I replicated John’s Census results from 2001-2016 for Vancouver, I apparently did not replicate his results for other metro areas. I admit, I didn’t catch this, since I was focused on Vancouver (and since I ran the replication of his 1.19 ratio of new dwellings to new households very quickly, before he’d provided his full report). I don’t know why his results and my own differ for metro areas beyond Vancouver, but it’s worth looking into! The data should be from the same source (Statistics Canada), but sometimes they report things differently in different documents, and it’s also entirely possible that errors were introduced in transcribing data (in which case, they were probably mine! My response was hastily assembled). Though it does not change the results for Vancouver, it’d be good to nail down overall dwelling count and occupancy changes.
  2. As John notes, the Census does not offer guidance with respect to how their procedural changes affect underlying dwelling count data between 2001 and 2006. But in noting their newly inclusive criteria for expanding the count of secondary suites, they clearly point out how single-family dwellings changed to duplexes in their structure data. This implies that each of those dwellings formerly counted as one unit (but containing a secondary suite) would henceforth be counted as two or more. As we know, Vancouver has a LOT of secondary suites, and this shift in classification both could and should have boosted the count of dwelling units significantly, even without any new dwellings being built or added. Worth noting as well that new secondary suites are the LEAST likely to show up in permitting data (though the Metro Van databook for 2017 at least tries to capture them). It would be great to get more from the Census on the characteristics of “dwellings unoccupied by usual residents.” On a related note: I’d love it if someone could point me toward or carry out an intensive study of how the Census counts dwellings in Canada!
  3. John acknowledges the awesomeness of the construction permitting data, but does not (yet) engage with how much better it fits new household formation than census counts of dwellings, indicating a shortage rather than a surplus of supply. I’ll look forward to seeing his comparison between construction and census data if he’s able to pull one together! (Both of us have time constraints involving stacks of grading and lots of other work on our plates).

Otherwise, as I said, John Rose is both a nice guy and clearly well-intentioned. He mentioned during our conversation that his study was motivated over concerns about new construction in the Agricultural Land Reserve (ALR) in Richmond. On this point, we clearly agree. The ALR is worth saving, and we don’t need to expand our housing supply any further out into Vancouver’s agricultural and wild lands, which is part of why I focus on densifying single-family residential neighbourhoods as the best path toward making Vancouver a more affordable, more inclusive, more lively, and more sustainable city.

 

[Postscript, Dec 15th: for more see Jens’ careful response with a detailed dive into the data over at MountainMath, and see the smart historical commentary on the Census in Vancouver in the comments below by the folks at Changing City (added bonus: see their lovely pictures of changing streetscapes around town!)

3 thoughts on “A Limerick (and a reply to a response to a critique of a study)

  1. That’s good that Rose is nice guy and well-intentioned.

    It’s a shame that he ended up sharing his work by way of Kerry Gold, who’s article pre-emptively poisoned the well with her suggestion that people who might be critical of Rose’s work were probably shills or bullies.

    Like

  2. When it first appeared I noted elsewhere (on Pricetags) that Dr. Rose didn’t seem to be particularly familiar with Census data, or how the Census collection has changed over time. As a result his conclusions were dubious. You seem to have reached pretty much the same conclusion.

    Citing change over 15 years, (so going back to the 2001 Census for his base data), has an inherent problem. That year Statistics Canada counted 786,285 dwellings in Metro Vancouver (technically, the Vancouver CMA), of which 758,720 were not occupied by ‘usual residents’. That’s a 3.5% rate of ‘not occupied by usual residents’. It’s actually less than in 1996, when 4.3% of dwellings were not occupied by ‘usual residents’.

    Note that these aren’t necessarily actually ‘vacant’ dwellings – they might be occupied by foreign or temporary residents, who don’t get counted. In 2001 6,735 dwellings were occupied on that basis but not considered occupied by usual residents. Other dwellings might be used as second homes, during the week by business people for example, but not considered ‘occupied’ because another dwelling is the family home. (Have you seen Jim Treliving in his Coal Harbour penthouse on Dragon’s Den?)There are a whole number of other reasons why a dwelling might not be counted as ‘occupied’. Some dwellings will be vacant because they’re being renovated, awaiting demolition, or for sale by owners who have already moved elsewhere. Some will be vacant because they’re in a recently completed apartment building, but not all the units have been occupied on the day of the Census.

    In 2006 Statistics Canada switched from collecting Census forms, to mailed and online Census returns. In making that change, they made a real effort to collect addresses to contact the population. One outcome of that change was that they found thousand of new dwellings, many of them secondary and basement suites. They weren’t necessarily ‘legal’ or authorized, but they collected addresses from all sorts of sources. If a bank statement or Cable TV bill was sent to ‘Basement Suite, 123 Wherever Road’, then they counted that as two dwellings in the home, and classified it as a duplex. You can see the change very clearly in Vancouver; the number of occupied single family dwellings dropped between 2001 and 2006 by nearly 40,000. Nobody thinks they were all demolished, or replaced in those five years – it’s just that most were now either a duplex (house with a suite), or if they found two suites, apartments under 5 storeys.

    In 2006 the proportion of dwellings that were ‘not occupied by usual residents’ went up sharply, to 6.2% of the dwellings. In 2011 it was almost the same, at 6.1%. In 2016, in the latest census, it went up a bit more to 6.5%. So over three census periods it really didn’t change very much. If the analysis is carried out using those three data points, the conclusion would be quite different.

    In the City of Vancouver there was a lot of fuss about how ’empty’ Marine Gateway and the area near Joyce station were in the 2016 Census. Some commentators assumed they were all being bought by investors and left empty, or flipped. But remember that some units will be vacant because they’re in a recently completed apartment building, but not all the units have been occupied on the day of the Census? Looking at the individual block counts of dwellings that’s exactly what happened in those two locations – there were several hundred units that were counted as existing – although they might not actually have had an occupancy permit – but nobody had moved in on Census day. That circumstance particularly applies when there’s been a building boom and large residential projects are coming on stream. That was more true in 2016 than it was in 2011 – there were over 18,000 dwellings completed in 2016 in Metro Vancouver, compared to 13,000 in 2011.

    There’s some suggestion that the number of dwellings occupied by temporary and foreign workers was also higher in 2016, but Statistics Canada don’t seem to have published those numbers yet. Remember those count as ‘not occupied by usual residents’ as well. As we also know, in 2016 some would have been Air B&B units as well.

    The 6.5% ‘not occupied by usual residents’ in Greater Vancouver is less than in BC as whole (8.8%), Canada as a whole (8.8%), less than Portland (7.8%) and similar to Seattle (6.4%). Edmonton and Regina have higher proportions; Montreal and Toronto a little lower.

    Statistics Canada had changed definitions of dwellings between 1996 and 2001, and the statistics geeks of the world were very unhappy. The sort of people who put the ‘anal’ into analysis couldn’t really compare data by dwelling type over the two periods. When there were further changes in 2006, Statistics Canada seem to have downplayed it, presumably to avoid more flack – although they did note the changes.

    There are other definitional changes that complicate a time series. Between 2006 and 2011, perhaps because getting mail-in responses had been difficult, in the City of Vancouver Statistics Canada switched a lot of ‘apartments’ to collective dwellings – mostly seniors homes and SRO hotels. That meant the number of occupied low rise apartments in the City for example apparently fell – which is patently inaccurate. Of course the occupancy of those SROs removed from the apartment stock would be close to 100%, so the ‘vacancy’ rate was also slightly affected.

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s