Data spotlight: GDP revisions, round 2
GDP growth tends to be revised up over time. One reason may be that the initial estimates haven’t been good at capturing population growth.
Earlier this year I wrote a piece looking at how GDP, our premier measure of the state of the economy, can end up being revised for several years after its initial release. In part this was prompted by the annual benchmarking revisions, which are published each December and tend to be when the largest revisions occur. Last year’s revisions were to the downside, suggesting that since the end of the Covid lockdown, the economy had grown by 1.4% less than previously thought.
Data spotlight: GDP revisions
Improved estimates suggest that the economy has grown by less than we thought. Is there anything more to the story?
Yesterday Stats NZ provided a preview of what this year’s annual revisions will look like. We don’t have the full results yet - they’ll be published along with the September quarter GDP report on the 19th of December - but it appears this time the revisions will be substantially upward. Growth for the year to March 2023 is expected to lift from 2.7% to 3.5% - more than reversing the downward revision that was made to this same figure a year ago. And for the year to March 2024, growth is estimated to rise from 0.3% to 1.4%.
In my earlier piece I listed a few reasons why the GDP figures can (and need to) be revised after their initial release. That description was fairly neutral and uncontroversial. But this time I want to put forward another potential reason that’s much more speculative, and no doubt a gross oversimplification of how GDP is calculated. But it crops up so regularly that I don’t think it can be ignored. The gist of it is: the quarterly GDP estimates don’t keep up with population growth very well.
As I discussed last time, the economy is far too big to be able to measure all activity on a regular basis. Instead, a lot of the information that goes into GDP comes from surveying a small share of businesses in each industry (which requires a lot of work to try to ensure that these samples are representative of the whole). The businesses that are surveyed are selected from a pre-determined ‘frame’.
This should give us a pretty good idea of how much growth there has been within the frame. My guess, however, is that it’s not so good at capturing whether the frame itself should be growing.
This will be even more of an issue when there’s a big swing in the rate of population growth, as often happens in New Zealand. (I’m fudging things a bit here; my supposition is that more people results in more businesses, rather than larger ones on average.) Part of the annual benchmarking exercise for GDP involves taking detailed data on the shape of the economy, which can be a few years old by that point, and rolling it forward into the last few years. So if your detailed data comes from a year of low population growth, and you apply it to years when population growth was high, it’s quite possible that you’ll miss something about how the economy has changed.
As it happens, the recent swing in New Zealand’s population growth rate has been unprecedented (or at least since the end of WWII). During the Covid border closure, net migration went from strongly positive to slightly negative, reducing overall population growth to around 0.5%yr. Then as the border was reopened, catch-up movements saw net migration surge to record highs, lifting population growth to around 3%yr.
What I think ended up happening is that we correctly recorded the increase in the number of people, but not the increase in the scope of the economy that came with them. Putting these figures together then left us with a picture of a steep fall in per-person GDP and labour productivity. Yesterday’s GDP revisions don’t completely overturn that story, but it looks much less worrying now than it did before.
As I said, a lot of this is speculation on my part. But there is one fascinating bit of evidence in its favour. If you look at the track record of GDP at its first release, compared to what it looks like today, there’s been a clear bias towards upward revisions. Since the beginning of the current GDP series in 1987, the rate of growth has been revised up from its first estimate by an average of around 0.5% per year. It seems entirely plausible to me that a persistent factor like population growth could be responsible for a persistent underestimation of GDP growth.
Since 1987, the rate of GDP growth has been revised up from its first estimate by an average of around 0.5% per year.
The thing that really prompted me into writing my previous piece was a speech by the Reserve Bank’s chief economist, which tried to run a story about how lower GDP actually pointed to more inflation pressure in the economy than previously thought. I was sceptical of that interpretation at the time, and it would be fair to say that it didn’t age well.
So what will the RBNZ make of these latest revisions? Can we expect a speech next year arguing the opposite side, that higher GDP growth actually means less inflation?
My hope is that they would just leave the issue well enough alone. These revisions are really a matter of good data hygiene; they’re about correcting the historic record, rather than telling us something new about the current situation. As I’ve said before, if you want to gauge how hot or cold the economy is running, just look at the unemployment rate and don’t overthink it.