International Visitor Survey Methodology
COVID-19 Methodology Changes
Due to COVID-19 there was no International Visitor Survey (IVS) interviewing in June quarter 2020. As a result some estimates for this period of the year ending June 2020 were imputed.
Visitor estimates in the IVS are benchmarked to Overseas Arrivals and Departures (OAD) short term visitor arrivals aged 15 years and over.
The OAD is currently using a full count (rather than sample) of passenger cards to measure short-term visitor arrivals to Australia due to the small number of visitor movements occurring. This more comprehensive OAD data has been used for IVS imputation purposes
OAD data indicates there were nearly 10,000 international visitors aged 15 years and over to Australia in June quarter 2020.
Using advanced statistical techniques, TRA took trip characteristics from relevant past IVS records with the same profile to each of the visitors in the OAD to impute values for a range of visitor characteristics in June quarter 2020.
Spend values have been sensitivity checked against domestic spend behaviours by item to ensure alignment with current spending patterns.
Changes to passenger movement data has led to revisions to International Visitor Survey data
TRA has revised International Visitor Survey (IVS) estimates from 2005 to 2018 with the release of year ending December 2018 data. The revisions focus on purpose-of-visit estimates, with minimal change to visitor, nights and spend estimates. As results have been revised back to 2005, please do not use/compare previously published data.
TRA uses data from the passenger processing systems supplied by the Department of Home Affairs to benchmark and produce a purpose-of-visit measure for the IVS.
In July 2017, Home Affairs discontinued the outgoing passenger card and changed the format of the incoming passenger card to be able to collect purpose-of-visit data for returning Australian residents. The changes to the card also affected the continued collection of main reason for journey for international visitors arriving in Australia. It soon became evident that the distribution of responses to the question on main reason for journey for short-term visitor arrivals was skewed, showing an increase in the number of persons reporting visiting friends and relatives (VFR) and a corresponding decline in the number of persons reporting holiday.
Investigations showed that anomalies existed in coding of purpose-of-visit data:
- Prior to July 2017, whenever passengers marked both ‘holiday’ and ‘VFR’ as their main reason for travel, records were coded to ‘holiday’, which over estimated holiday visitors and under estimated VFR visitors.
- The opposite occurred from July 2017 onwards, where records were coded to ‘VFR’ when both ‘holiday’ and ‘VFR’ were marked, resulting in holiday visitors being under estimated and VFR visitors being over estimated.
These anomalies caused large shifts in purpose-of-visit records between these periods.
In response, the ABS and Home Affairs revised imputation and processing practices of the passenger card data, and made a number of changes that included:
- The new system allocating data based on traveller characteristics, which improves the allocation of records where two main purpose-of-visit categories are marked, i.e. ‘holiday’ and ‘VFR’.
- Addressing an issue where travel agents pre-printed passenger cards with the old card design, causing a small number of purpose-of-visit records to be allocated incorrectly.
The International Visitor Survey (IVS) samples 40,000 departing, short-term international travellers aged 15 years and over who have been visiting Australia. The survey is conducted by Computer Assisted Personal Interviewing (CAPI) in the departure lounges of the eight major international airports: Sydney, Melbourne, Brisbane, Cairns, Perth, Adelaide, Darwin and the Gold Coast.
The IVS contains around 100 questions supported by 'show-cards' that are used to help the respondent answer particular sections including:
Usual place of residence
Sources for obtaining information about Australia
Purpose of visit and places visited
Transportation and accommodation
The survey design and management is the responsibility of the National
Survey Section in Tourism Research Australia (TRA). The section works
closely with the consultants, key stakeholders and industry to develop and
maintain high data quality and relevant outputs.
Since 2004, the IVS has been surveying international visitors in four
languages: English, Japanese, Mandarin and Korean. The total number of
interviews conducted with particular residents of each country or region is
distributed among airports by selecting monthly samples of departing
flights and visitors on those flights to achieve acceptable sample sizes in
Survey results are weighted to data on international visitor numbers over
the period, provided by the Department of Immigration and Citizenship
(DIAC), with the assistance of the Australian Bureau of Statistics (ABS).
The variables used in weighting the data are:
country of residence
state of arrival
main purpose of journey
airport of departure
age and sex of visitor.
Overseas (visitor) arrivals and departures (OAD) data are also published by the ABS (ABS catalogue no. 3401.0) on a monthly basis.
Increase in sample size for 2005 survey
Between 2001 and 2004 interviews were conducted with approximately 20,000
international visitors aged 15 years and over as they were departing
Australia. Since 1 January 2005, interviews have been conducted with 40,000
international visitors on an annual basis. The sample was increased in
order to enhance the estimates for smaller states, territories and regions.
Increasing the sample size of the IVS by 100% has improved the reliability
of survey estimates.
Visitor interview by country or region of residence
The table below shows the number of interviews conducted in the December Quarter 2012 and for the year ended 31 December 2012.
Sample size by country or residence
The results given in the IVS are based on a sample, rather than a census,
of international visitors to Australia. As with all sample surveys, the
results are subject to sampling variability, and therefore may differ from
figures that would be obtained if all international visitors to Australia
had been included in the survey.
A measure of the possible degree of difference is given by the relative
standard error of the survey and its associated confidence interval, which
indicates the extent to which an estimate might vary by chance from the
true figure because only a sample of the population was included.
The table below provides the 95% confidence interval widths for a range of
estimates available in the IVS. That is, there are approximately 19 chances
in 20 that the true number is within the range identified by applying the
figures in the table.
Size of 95% Confidence Interval for Estimate (expressed as a percentage of the estimate)
| 1,000 000
| 2,000 000
| 5,000 000
| 10 000, 000
| 20 000, 000
| 50 000, 000
| 100 000, 000
| 200 000, 000
| 500 000, 000
| 100 000, 000
| 200 000, 000
| 500 000, 000
| 10 000 000 000
# - 95% Confidence Interval is greater than estimate.
The following example illustrates the use of this table to determine a
range within which we are 95% confident that the true total lies. Say, the
estimated number of Chinese visitors who stayed in Queensland was 100,000.
Looking at the visits column (see table), an estimate of 100,000 visitors
has a 95% Confidence Interval of 9.9%. Thus we are 95% confident that the
true number of Chinese who stayed in Queensland was between 90,100 and
109,900 visitors (100,000 ± 9.9%).
The IVS relative standard errors were calculated using the Complex Survey
Sampling module in SPSS V14.0. Estimates of variation are based on sampling
with replacement principles and makes allowances for the IVS
stratification. The covariance and estimates output from this program were
then regressed with a log transformation using Ordinary Least Squares (OLS)
regression to achieve three independent models (for visitors, nights and
expenditure). The models were computed using the R statistical program and
the actual relationship modelled was:
1n (COV) = a + b * 1n (ESTIMATE)
Where, a = intercept, b = gradient (slope)
The model parameters were approximated as: