Why are Auckland housing consents falling?

Here's a rather worrying graph. It's the number of consents each month for new dwelling units in the Auckland region, on a 'trend' basis (the statisticians' best effort to remove seasonal variation and random noise).

I've put up the whole history of this series*, back to the start of 1995, because it tells us several interesting things. One is that, while people rightly point to a variety of reasons for Auckland's current housing shortage, one that tends to get forgotten is the GFC: at a rough estimate, at least 7,500 houses didn't get built in that period that would normally have been. And another is that despite the recent strong rise in consents, we're still not up to the best levels of the past (in the early 2000s), even though the need is much more pressing these days.

The thing I wanted to highlight, though, is the drop-off at the right-hand end.The number of consents has been falling since September and October - it was 812 in both those months - and has since dropped to 766 (in March, the latest date available).

It's certainly a very odd development. It seems at variance with what you see on the ground in Auckland. Sure, there are lags between consents and construction, and a lot of the current activity could be down to the higher levels of consents issued earlier last year, but if there was a genuine drop-off in the housing consent pipeline over the past six months, you'd think you'd be seeing some slackening off in actual construction activity by now. And you're not, at least on my subjective 'economics by walking around' assessment. Everywhere I go around Auckland's North Shore, there are new developments I hadn't seen before.

Given that the data, at least to me, don't line up with the reality I see, I've been doing a bit of tyre-kicking with the ever-helpful Statistics NZ staff, on this occasion Danielle Barwick in the Christchurch office (and I should say all views here are mine, not hers).

My first thought was perhaps the number of dwelling units per consent might have been going up. Ten years ago, perhaps the 'typical' consent was for one detached house: maybe, these days, one consent could be for 20 terraced townhouses? Some of the new developments are high-density indeed: here's one I snapped today at Silver Moon Road in Albany.

So consents, perhaps, could be going down but with increasing dwellings per consent, dwellings could still be going up? Nope. A single consent for a 30 house development gets counted as 30 dwelling units.

There is another possibility, though. At the early stages of a project, there can be a consent for the earthworks part of a project: at that stage the final shape of the development isn't yet known (so Stats can't put a dwellings unit number on it, though they will when it's finalised). You could have a single consent actually representing a very large project until the full size gets logged later. And if there are a lot more of those happening - and there are lots underway, here's another local one, off Spencer Road, again in Albany - then the consents numbers could be temporarily lower but the pipeline could actually be getting larger.

Maybe that's part of the answer, and we'll all be relieved when these earthworks-stage projects eventually get counted at their full value. Maybe something has fried the brain of the trend-identifying algorithm, and it'll all come right with a few more months' data. Maybe the planners have been experiencing some kind of consent back-up or logjam (though they've been handling higher numbers in the past, so why now?). And it could be that this recent drop-off in consents is just one of those things: there have been fluctuations in the past that went against the longer-term trend for a while before getting back on track.

At the moment, though, it looks distinctly odd. And if it keeps up for any further length of time, it will go from 'distinctly odd' to 'downright alarming'.

*You can find it for yourself, if you're minded, on Stats' Infoshare service, search for the series identifier BLDM.SF021100A1T