It will vary by provider, but I use SQM research.
Their methodology is explained here as
http://www.sqmresearch.com.au/graph_vacancy.php?postcode=2000&t=1#terms
Rental Vacancies
The Rental Vacancies component is based on all monitored and unique online listings for the period of a calendar month. The series starts off in January 2005.
All listings are taken from online monitoring of major listings sites. Only those properties with unique addresses or a unique listing id are used. Those advertisements with no addresses are excluded from the series. Any addresses repeated between sites are de-duped.
Only those listings that have been advertised for three weeks or more (and are still currently advertised as at the time of collation) are used.
Established Dwellings
We use the 2001, 2006 and 2011 No. of total established Dwellings (as a base) by postcode as determined by the ABS census. In addition we estimate total dwellings for 2005, 2007, 2008, 2009, 2010 and 2012. We then multiply this by the percentage of renters for each postcode as also provided in the census. This provides an estimated available total stock for rent.
The numerator is then divided into the denominator, which provides a vacancy rate percentage.
Some have argued that using online real estate listings cannot be done because of advertising of false listings and properties that are only advertised for a fleeting moment as they are taken up immediately. We have addressed those issues.
We have left out those ads without an advertised address or unique property id and have also taken into account a whole month’s worth of listings which automatically adjusts for those ads withdrawn quickly.
