Can we trust our migration statistics?
Monday, 14 August 2023
Migration figures due out on Monday are expected to show New Zealand citizens leaving the country but continuing to be replaced by a larger influx of new arrivals.
Swings in immigration and emigration affect the New Zealand economy more than most.
But they can be hard to measure precisely, at least while they are still fresh.
Reserve Bank deputy governor Christian Hawkesby noted in May that the official migration statistics produced by Stats NZ are often later revised and said the central bank hadn’t always taken them at face value.
But the key is to understand what exactly is being measured, as it may not be what you think.
So who is a migrant?
Stats NZ counts someone as a migrant to New Zealand if it thinks they are going to spend at least 12 months out of the next 16 months living here.
Similarly, it classifies someone as having emigrated if it thinks they are going to spend more than 12 out of the next 16 months out of the country.
It is the same definition Australia uses and dovetails with the United Nations’ guidelines on population statistics.
But it does mean a lot of people who might not fit the traditional image of an immigrant or emigrant could temporarily qualify, such as people who are on working holidays here or overseas, or students travelling here or overseas to study.
Why is migration hard to measure?
Being an island, it is easy to count the number people crossing the New Zealand border.
What is harder is working out whether and when visitors might leave, and when people leaving the country might come back.
That is especially the case for New Zealand citizens and residents, who are free to come and go as they please.
They are a very important group in the migration statistics, as the biggest swings in migration tend to result from more New Zealanders leaving for Australia, or from more New Zealanders returning home from overseas.
The number of foreign citizens arriving in New Zealand to make it their home is more constant.
So how does Stats NZ measure migration?
It used to base its numbers on the information people gave on their passenger arrival and departure cards about where they expected to live for the next 12 months.
But that approach had to change when the Government scrapped departure cards in 2018.
It had some drawbacks anyway, as people’s plans can and do change.
A visitor intending to study in New Zealand for six months or a Kiwi returning from overseas to visit their family might end up settling here, for example.
Similarly, a New Zealander heading overseas for short break might decide to stay away longer than they expected.
One upshot of the old approach was that the official figures tended to understate both migrant arrivals and departures.
And now?
Stats NZ makes its own predictions using a machine-learning tool that assesses how likely it is people who have crossed the border have come to live here for a year, or have left the country for the longer term.
It does this based partly on how long they have stayed or been away, which eventually becomes a clear measure of whether they have migrated or not.
But in the interim, it also bases its guess on the chances of people being migrants based on considerations such as the type of visa people have travelled on, their country of citizenship, and even their age.
How accurate is the data?
There is always going to be a trade-off between the migration statistics being prompt and being accurate, and an estimate that may be a bit rough is usually going to be more useful than one that’s outdated.
Initially, the monthly estimates may not necessarily be very accurate, but then they will become more and more accurate over time, until 17 months after the figures are first released when they should be rock solid.
By way of example, Stats NZ estimated in November that the country had a net inflow of 2176 migrants in September 2022, but by May this year it had revised that figure up to 4704, so a decent difference in that particular case.
The September estimate is still bouncing around a bit, but each month the revisions should get smaller, and the annual estimates will be more reliable.
It is also the case that Stats NZ’s machine-learning model should get better at making its estimates over time, as it learns from more data and more results.
Most caution may be needed around the numbers when there is something unusual going on that could trip up the model – like a global pandemic.