February 08, 2019

Migration measures - timely or accurate?

Statistics New Zealand’s recent change in methodology for reporting migration data has brought to the fore the debate about providing timely data or accurate data.

When Statistics New Zealand announced its new approach to measuring migration, with the removal of departure cards for people leaving New Zealand from November 2018, BERL stated that researchers and policy makers need to step back and ask:

How important is timely migration data?

Will Statistics New Zealand’s new model be able to identify changes in migration trends?

And, is the removal of the burden on travellers (departure cards) and reduced processing costs worth the potential lessening in the robustness of a timely migration measure?

With the release of migration numbers in November 2018 using Statistics New Zealand’s new methodology, researchers and policy makers now need to answer these questions, and their answers will determine which migration measures they will use going forward.

This is because the release of migration numbers for November 2018 has seen Statistics New Zealand apply its new methodology, which tries to be both accurate and timely - using two different measures.

The first measure classed anyone who stayed in New Zealand for at least 12 months across a 16 month period as a permanent migrant while the remainder were classed as short-term visitors. This approach should be very accurate, using actual Customs New Zealand records. The main drawback of this measure is that Statistics New Zealand is only able to produce statistics 17 months in arrears, and therefore not a timely measure.

The second measure uses a probabilistic predictive model of travel type (short-term arrival or permanent migrant) based on the characteristics of travellers. Whilst this measure is timely, it is not necessarily that accurate, reflected in the fact that Statistics New Zealand publishes an estimated standard error for the measure.

In other words, the latest releases of official migration data, based on the new “probabilistic predictive model”, will likely be revised 16 months later to incorporate the actual Customs records. As an indication of our concern, the following graph shows a sudden rapid upturn in departures for the November 2018 month. This appears at odds with the previous movements in the ‘old series’ measure and Statistics New Zealand’s own estimates for previous months, which showed small increases in previous November months.

Across previous months the new series for departures showed fairly similar movements across the last two years to the old series, though we do see high spike in December months. This strong upturn, after a few months of strong declines in the “old series” does unfortunately bring into question the validity and usefulness of the Statistics New Zealand model, and its ability to predict timely changes rather than repeat the trends of the last few years. Unfortunately this question on the validity of the strong increase in November 2018, will not be satisfactory answered for another 17 months, when the final numbers for November 2018 are determined, and we can finally see how accurate the modelled movement was.