In the January Raving I suggested Statistics NZ may have botched the new migration numbers. Large revisions to the emigration numbers when the December numbers were released today give support to my concern.
But there may be more reasons to question the numbers produced by the statistical model Stats NZ has built to fill the gap left by the dumping of departure cards.
Even after a friendly call from Stats NZ in response to the last Raving I am still sceptical of the statistical model used to estimate numbers for the last 16 months; but Stats NZ fixed most of the problems I identified with Infoshare.
Revised emigration numbers look more realistic, supporting the concerns raised
The chart below shows emigration numbers based on the cards that were dumped in November 2018 (black line), the estimated numbers released last month that I suggested were wrong (green line) and the new estimates released today (blue line). The chart shows annualised, seasonally adjusted, threemonth averages. The new annualised estimate for the three months ended November is 15,360 lower than the estimate released last month. Ouch!
The gap in late-2018 between the estimated numbers and the numbers based on departure has narrowed hugely since last month. It may narrow more as actual numbers in time replace the estimates produced by the statistical model. See the last Raving for a discussion of the statistical model and my concerns about it.
Something is fishy with the immigration numbers but there could be two reasons
In the last Raving I raised some concerns about the immigration numbers but last month they didn't look as problematic as the emigration ones. However, the chart below suggests something very fishy is now going on with the immigration numbers.
The previous chart uses annualised, seasonally adjusted, three-month averages. The black line is based on the cards that are still in use and suggest immigration numbers are plunging. The green line shows the estimated numbers released last month up to November 2018 and the blue line shows the new estimates released today. There isn't a huge gap between the new numbers and those released last month, but both differ hugely from the cardbased numbers.
There are two options but it may be a bit of both:
• A massive fall in immigration is underway and the statistical model is no good at picking turning points.
• A larger number of immigrants arriving on a permanent or long-term basis have taken to filling in the arrival cards wrong.
Last month commonsense told me that the new estimated emigration numbers were wrong. The huge revisions to the numbers give credence to my concern. Commonsense now suggests something is dramatically wrong with the immigration numbers (i.e. the yawning gap that has opened up in the previous chart between the card-based numbers and the model-based estimates is beyond the realm of reasonable). I am still sceptical of the surge in immigration suggested by the statistical model estimates but I am also sceptical of the plunge suggested by the cards. In time actuals will replace the estimates, revealing what actually occurred. But why hasn't Stats NZ highlighted and tried to explain the massive discrepancy?!
Stats NZ has largely fixed the Infoshare problems identified in the last Raving
The head boffin at Stats NZ kindly called me in response to the last Raving to discuss the issues I raised. He acknowledged the problems identified with Infoshare regarding it containing several different measures for the immigration numbers. The problems were quickly and largely fixed, but more could still be done to identify which series were discontinued due to the dumping of the departure cards in November 2018.
I reiterate my warning to be wary of the new estimated migration numbers
After reviewing the latest numbers I reiterate my warning to clients and others to be wary of the statistical model's estimates. There will probably be more major revisions. The model may not be good at picking turning points. It is currently picking an upturn in immigration when it is possible a major downturn has recently started.