Petit Poll uses biased samples. Traditionally, this would make our results meaningless, but we are able to make reasonable conclusions because we tie our samples to representative data.
The best option for this representative data would be a large exit poll from the 2013 election. But, I was not able to find an appropriate dataset. Luckily, as an ANU student, I have access to the Australian Election Study (AES) database. This is available via the Australian Data Archive - thank you! By way of background, the AES coincides with Federal election and has been going on since 1987.
On review, I found out that our initial database had not been weighted. So our underlying dataset did not reflect the Australian population to its fullest extent. I’ve since downloaded the weighted database, and this now underpins our model.
Unfortunately there are restrictions with what we are able to use this database for. It is also reasonably small. A better solution would be to partner with an organization that has a large exit poll dataset that they are willing to make available. Anyone know someone?