Design and Estimation for the Agricultural Land and Water Ownership Survey
Readers may be aware that foreign ownership of agricultural land in Australia has been the subject of considerable media and community interest in recent months. When this issue was raised in Parliament late last year, Treasury officers approached the ABS with a request to conduct a survey to inform the Government on the degree and extent of foreign ownership of agricultural businesses, land and water. This request was passed on to the Environment and Agriculture Business Statistics Centre (BSC) who were asked to commit to a release date of September 2011. Anyone with experience of survey development and conduct will appreciate that this is a very tight timetable, especially in a year when the BSC is conducting the five-yearly Agricultural Census.
The outputs desired are business counts, land area and volume of water entitlement, by degree of foreign ownership of the business (5 categories) and at national and state level. These are to be produced from a sample of about 10,000 from a frame of 162,000 units. The survey presented the BSC with a number of challenges from the outset:
- the degree of foreign ownership can be difficult to determine where there are a number of businesses in the "ownership chain", or in cases of joint ownership
- there are regional differences in the way water use is regulated
- the frame used (the preliminary frame for the Agricultural Census) comprises all Australian-registered businesses with known agricultural activity. This is the only practical frame that could be used, but requires us to overcome the challenge posed by farms that lease land from non-agricultural businesses which are not on the frame.
Sample design also had a number of challenges:
- there is little relevant data to guide sample allocation (a survey of foreign ownership in the agricultural sector was last conducted in 1984)
- the businesses of interest are a sparse sub-population (only about one in 160 businesses is expected to have some degree of foreign ownership)
- the only variable available for sizing units does not correlate very well with land area
Initial work by the BSC identified a small number of units thought to be foreign-owned and the distribution of these units over industry and geographical region was used to inform the stratification. There was no size component of the stratification other than a separation of units into the completely-enumerated and sampled sectors. Instead, the importance of larger units was captured by making the selection probability proportional to a measure of the unit's agricultural production. Unconditional Poisson selection was used to select the sample. Since this means that sample sizes are not fixed, weights were calibrated to post-stratum populations in order to improve the accuracy of estimates and variances.
The publication was released on September 9, the culmination of a significant effort by the BSC and its MDMD support to meet a non-standard request in a difficult time frame.
For more information, please contact Jos Beunen on (02) 9268 4647 or firstname.lastname@example.org