IVA Data Reliability
Last updated:
April 2008
Background
The International Travel and Migration data collection is a both a census and a survey. It is a census in the respect that it is compulsory to fill out the arrival and departure cards from which arrivals & departures data is derived.
Some data are captured 100% such as permanent and long-term travelers, and some short-term traveller variables (age, sex, country of visitors, NZ port and citizenship) and therefore have negligible sampling errors. However, some short-term traveller variables that are sampled are subject to sample errors including:
- Reason for travel
- Intended length of stay/absence
- Country of birth
- Overseas port of embarkation/disembarkation
- Overseas state/province of residence
- Country of main destination
Sampling errors measure the variability that arises because the entire population has not been surveyed. They are calculated at a particular confidence level - in this case 95%. Thus we are 95% certain that the true population value falls between the sample estimate plus the absolute sampling error, and the sample estimate minus the absolute sampling error. Relative sample errors give a percentage measure of the magnitude of the error, obtained by dividing the absolute sample error by the size of the cell of interest.
Each visitor arrival card is classified by the country of last permanent residence (CLPR). Sampling of cards is done within each country.
The sampling fraction varies by month, tending to be higher in winter and lower in summer. For example, in January 1998, the sampling fraction was 1 in 15. This means that every 15th card which had a CLPR of Australia was sampled, as was every 15th card which had a CLPR of Hong Kong etc.
For more detail on sampling error for International Arrivals and Departures can be found in the the document below.
International Visitor Arrivals/Departures Sampling Error (87KB)
Non-Sampling Errors
It should be recognised that any estimates will also have a non-sampling error component, which is generally unmeasurable. In the case of migration statistics, non-sampling error could result from factors such as:
- non response
- incorrect responses due to language difficulties
- processing errors
- a person changing their travel intentions after completing the migration card
In the preparation of migration statistics, the classification of each passenger is primarily determined by the passenger's response to the question on intended or actual length of stay/absence. If the person's intention changes later during the trip, then he or she may also change their migrant category, e.g. a person may come to New Zealand with the declared intention of settling permanently, but in fact may return overseas after a few months.
In 1995 New Zealand Customs and the New Zealand Immigration Service negotiated an agreement covering the collection of arrival and departure cards. The agreement specified that Customs were to check only certain questions on the cards - principally passport and flight numbers. Previously, Customs had checked the whole card. As a result of this, non response rates increased between 1995 and 1996 for a number of variables.
Source: Statistics New Zealand
ˆ Back to top