Stats NZ has a new website.

For new releases go to

www.stats.govt.nz

As we transition to our new site, you'll still find some Stats NZ information here on this archive site.

  • Share this page to Facebook
  • Share this page to Twitter
  • Share this page to Google+
Alternative methods for measuring permanent and long-term migration

This article describes Statistics NZ's current method of recording net permanent and long-term migration based on passenger intentions, and examines some alternative methods based on passenger behaviours.

  • Statistics NZ regularly publishes permanent and long-term (PLT) migration statistics, which are determined from responses on arrival and departure cards.
  • A limitation of these statistics is that they are based on passenger intentions. People may change their intentions, but the published statistics are not revised to take account of these changes.
  • We have investigated several alternative methods of measuring PLT migration that are based on actual behaviour rather than intentions. These methods have not yet been developed to the point where they are ready for release as official series. Nonetheless, it can be useful to consider how the estimates of PLT migration vary across the different methods, and how these compare with the published statistics based on intentions.
  • The pattern of PLT migration over time is similar across all methods, but there are some differences in the volume of PLT migration across the different methods.
  • All the alternative methods have limitations that require further investigation. The official PLT migration series based on passenger intentions remains the best currently available estimate of PLT migration.

Current permanent and long-term migration series

Statistics NZ publishes monthly permanent and long-term (PLT) migration statistics that we calculate from passenger intentions as stated on arrival and departure cards.

People are classified as PLT arrivals if they have been living overseas for the last 12 months or more, and state they intend to live in New Zealand for the next 12 months or more.

People are classified as PLT departures if they have been living in New Zealand for the last 12 months or more, and state they intend to live outside New Zealand for the next 12 months or more.

Net PLT migration is calculated by subtracting PLT departures from PLT arrivals.

Those people not classified as PLT migrants are classified as one of two short-term traveller types. People who state they do not live in New Zealand and will be here for less than 12 months are classified as overseas visitors. People who state they live in New Zealand and will be away for less than 12 months are classified as New Zealand-resident travellers.

Accuracy of recorded passenger type

Actual passenger behaviour can differ from the recorded passenger type because people change their intentions, fill out arrival and departure cards incorrectly, or are classified incorrectly during processing. Four passenger behaviours can affect PLT migration statistics:

  1. Someone recorded as a PLT departure, who returns to New Zealand to live less than 12 months later. This person was actually a New Zealand-resident traveller.
  2. Someone recorded as a New Zealand-resident traveller departure (ie leaving New Zealand for less than 12 months), who ends up living overseas for 12 months or more. This person was actually a PLT departure.
  3. Someone recorded as a PLT arrival, who returns to live overseas less than 12 months later. This person was actually an overseas visitor.
  4. Someone recorded as an overseas visitor arrival, who ends up living in New Zealand for 12 months or more. This person was actually a PLT arrival.

These changes can counteract each other. For example, (1) decreases PLT departures, but (2) increases PLT departures. Furthermore, if the number of actual PLT arrivals is higher than recorded PLT arrivals, and the number of actual PLT departures is also higher than recorded PLT departures, there may be little effect on net migration (the same would apply if both arrivals and departures were lower than recorded totals).

Net passenger movements

A method that may, at first glance, appear to measure actual net PLT migration is all arrivals minus all departures (known as net passenger movements). Given that each short-term trip consists of an arrival and a departure, which cancel each other out to give a net total of zero, it is sometimes assumed that any net difference must be due to PLT migration. However, this assumption is not correct due to the following factors.

Short-term travellers in and out of New Zealand

The net passenger movements series is affected by changes in the numbers of short-term travellers in and out of New Zealand at either end of the time period.

For example, if 100,000 visitors are in New Zealand at the start of the period and 150,000 at the end, then 50,000 more visitors have arrived than departed over the period. This contributes 50,000 to net passenger movements. Changes in the number of New Zealand-resident travellers overseas similarly affect net passenger movements. Short-term travellers can have a big effect on net passenger movements, as they account for 98 percent of all passengers and can vary significantly in number over time.

Missed arrivals and departures

International travel and migration data has very high coverage, due to New Zealand’s robust border procedures. However, it is possible that a very small proportion of records are missed or double-counted. Some of these cases of ‘missed’ records are legitimate. The best example is flight crew who are on duty in one direction (and not counted in international travel and migration records because they are crew), but are ferried back off duty in the other direction (and are counted because they are passengers). Such a small percentage of records has little effect on key measures such as visitor arrivals and net permanent and long-term migration. However, the net passenger movements series is very sensitive to a small rate of error, given it is a very small difference between large flows (about 5 million arrivals and 5 million departures per year).

Comparing net PLT migration and net passenger movements

Figure 1 shows recorded PLT migration and net passenger movements for each June year from 2000 to 2013. The figure shows the net passenger movements series is highly volatile. For this reason, and the reasons outlined above, we do not recommend using the net passenger movements series as an estimate of PLT migration.

Figure 1

Graph, Comparison between net passenger movements and recorded net PLT migration, year ended June 2000 to 2013.

Alternative methods for measuring PLT migration

We have been exploring different methods for estimating PLT migration, with the aim of developing a method that does not rely on passenger intentions. Over time, we have developed several alternative methods. This section describes these methods.

End-points method

The Short-term travellers in and out of New Zealand section above demonstrates how changes in the number of short-term travellers contribute to net passenger movements. If we can estimate the number of short-term travellers in and out of New Zealand, we can remove the effect of these passengers from net passenger movements, leaving an estimate of permanent and long-term migration.

The end-points method estimates the number of short-term travellers in and out of New Zealand by identifying those people recorded as overseas visitors and New Zealand-resident travellers based on their arrival and departure card responses, and determining when they were in or out of New Zealand.

The limitations of this method are:

  • The end-points method relies on passenger type (resident, visitor, or PLT) assigned during processing, which can be incorrect. However, a person can only be incorrectly counted as a visitor in New Zealand or a New Zealand resident overseas for a short period. For example, someone who enters as a visitor, but changes their intentions while in New Zealand and stays for more than 12 months (so should actually have been recorded as a PLT arrival), will be counted as a visitor in New Zealand for the intended duration of their stay (which, by definition, must have been less than 12 months). After this period they will be correctly excluded from the estimate of visitors in New Zealand, and therefore will be correctly counted in the net migration estimates.
  • PLT migration estimates from the end-points method are not available until approximately 13 months after a given reference date.
  • Limited detail is available about who was actually a PLT migrant because the method works at the aggregate level rather than the individual level. Only net PLT migration, by age and sex, can be calculated.

The passport-linking method

This method aims to address one of the limitations of the end-points method by calculating passenger type (resident or visitor) using travel histories rather than relying on the passenger type assigned during processing.

Passenger type is assigned in two steps:

  1. Linking together individual passenger journeys to create a travel history for each person. This is done using passport number and date of birth, as passenger names were not available at the time that this method was developed.
  2. Calculating the proportion of time spent in New Zealand over the two years spanning (one year either side of) the reference date. People who have spent at least 50 percent of their time in New Zealand are classified as residents; all others are classified as visitors.

We then use these classifications to calculate New Zealand residents temporarily overseas, visitors in New Zealand, and then net PLT migration as in the end-points method.

The limitations of this method are:

  • Accurately linking journeys without passenger names is difficult, because some people travel on several different passports, or may change their passport between journeys. The method therefore relies on a number of assumptions to link individual passenger journeys, and these assumptions may not be correct, resulting in linking errors.
  • Many people do not have a continuous stay or absence of 12 months or more, making one or multiple short trips during a longer stay in or absence from New Zealand. Furthermore, some people have two consecutive arrivals or departures due to missing records. This method requires us to make several assumptions about the time spent in or out of New Zealand in order to classify these people. We haven’t yet tested the sensitivity of this method to changes in these assumptions.
  • PLT migration estimates from the passport linking method are not available until approximately 13 months after a given reference date.
  • Limited detail is available about who was actually a PLT migrant, because the method works at the aggregate level rather than the individual level. Only net PLT migration, by age and sex, can be calculated.

Linked-journeys method

The linked-journeys method uses passenger names as well as passport numbers and dates of birth to link passenger journeys together. This enables us to link journeys even when a person changes their passport. Personal identifying information (such as names and passport numbers) is securely stored in accordance with our standard confidentiality and privacy policies, and is not published. This method involves two steps:

  1. We link individual passenger journeys using name, passport number, and dates of birth to create a travel history for each individual. A match is made if 2 of the 3 variables match exactly.
  2. We calculate time spent in New Zealand during the 24 months before the reference date and the 24 months after the reference date. If someone spends more than 12 out of 24 months in New Zealand we classify them as a New Zealand resident, otherwise we classify them as a non-resident. If someone was a resident over the 24 months before the reference date, but not over the 24 months after the reference date, we classify their journey as a PLT departure (and vice versa for PLT arrivals).

Unlike the previous methods, this method calculates PLT arrivals and departures directly, rather than calculating net PLT migration by subtracting the effect of short-term travellers from net passenger movements. This method therefore has the potential to provide information about the characteristics of PLT migrants, although further development of the method will be required to do this.

The limitations of this method are:

  • Linking errors (which may occur if individuals travel under different passports with different names, if they change their name, or if there are errors in the recorded passport details) introduce error into the estimates of PLT migration. Initial evaluations of the linking in this method suggested the overall rate of linking errors is probably low, but that these errors are more likely to affect PLT migrants than other travellers.
  • Many people do not have a continuous stay or absence of 12 months or more, making one or more short trips during a longer stay in or absence from New Zealand. This method requires several assumptions about the length of time needed to be spent in or out of New Zealand to be considered a New Zealand resident and therefore to receive a PLT classification on arrival or departure. The sensitivity of this method to changes in these assumptions has not yet been tested.
  • If a person has multiple arrivals and departures, this method also requires assumptions about the timing of the PLT movement, ie which arrival or departure is the PLT journey.
  • PLT migration estimates from the linked journeys method are not available for some time after a given reference date. The length of delay varies depending on the assumptions made, but would likely be between 13 and 25 months.

Comparison of estimates from different methods

Figure 2 shows net PLT migration, by June year, as calculated by each of the methods. The figure shows the pattern of PLT migration is similar across all four methods, with the highest levels of net PLT migration recorded in 2002–03 and the lowest levels recorded in 2000, 2001, and 2012. However, the volume of net PLT migration varied across the different methods. For example, all methods show a peak in net PLT migration in 2002 and 2003, but the size of this peak is higher for the three alternative methods than for recorded PLT. From 2000 to 2009 the net PLT migration estimates from the alternative series were higher than the recorded statistics, but were lower over 2010–13.

Figure 2

Graph, Comparison between net recorded PLT and alternative net PLT migration methods, year ended June 2000 to 2013.

Conclusions

The official PLT migration series based on passenger intentions remains the best currently available estimate of PLT migration. While there are some differences in the volume of PLT between the different methods, all methods show a similar pattern over time, suggesting the published pattern of net migration is a good indication of highs, lows, increases, and decreases over time.

All the alternative methods have limitations that need further investigation before we could consider them for release as official series. Furthermore, all the alternative methods require a delay of at least 13 months before estimates are available, so the statistics based on passenger intentions remain the only option for the most recent year. We will continue investigating alternative methods as part of an ongoing continuous improvement programme.

More information

For more information, contact:
Sheree Gibb or Nicholas Thomson
Christchurch 03 964 8700
Email: info@stats.govt.nz

For general enquiries contact our Information Centre:
Phone: 0508 525 525 (toll-free in New Zealand)
+64 4 931 4600 (outside New Zealand)
Email: info@stats.govt.nz

ISSN: 1178-3877 (online)

Published 2 December 2014

  • Share this page to Facebook
  • Share this page to Twitter
  • Share this page to Google+
Top
  • Share this page to Facebook
  • Share this page to Twitter
  • Share this page to Google+