Since 2016 Petit Poll has used machine learning to make political polling that is meaningful but cheap. We combine data from biased surveys with statistical methods to generate meaningful results. We are able to gain an understanding of opinion at a more granular level because biased surveys have lower costs and our model shares information over time and space.

We gather non-representative samples combined with demographic and other data. We then use statistical models to ‘scale up’ the samples to provide results that are meaningful and akin to representative results. Petit Poll augments, rather than replaces, traditional polls such as Roy Morgan, Newspoll, Galaxy, etc. Petit Poll’s statistical approach is developed in Wang, Rothschild, Goel and Gelman (2014).

In contrast to many other Australian polls we focus on seat counts as the most direct determinant of who will win. For instance, as at the 2019 Federal Election, there were 151 seats up for grabs. If one party can get 77 seats then they have a clear majority government (one seat would be lost to appoint the speaker). Nation-wide two-party-preferred estimates are only an approximation to this. Also, our use of a model sets us apart from seat-level polling, as the model can take into account seat-specific issues. Statistical models perform better because they can learn from history, and take additional information into account. Get in touch if you’d like to learn more.


Monica and Rohan Alexander are the co-founders of Petit Poll.

Monica is an Assistant Professor in Statistical Sciences and Sociology at the University of Toronto. She received her PhD in Demography at the University of California, Berkeley. She has built these types of models for UNICEF and the WHO. Her research interests include statistical demography, mortality and health inequalities, and computational social science.

Rohan is a PhD student in the Research School of Economics at the ANU who expects to graduate in December 2019. He holds a Bachelor of Economics degree from the University of Queensland and a Master of Economics degree from the ANU. As part of a small team he co-founded GoCampaign which is now part of GO1, and has worked on political campaigns in Australia and abroad.

Rory Haupt is an analyst at Petit Poll. Rory is studying at the ANU, undertaking a Bachelor of Economics/Bachelor of Political Science double degree, with intentions of doing Honours in Political Science next year. Outside of his studies, Rory maintains a keen interest in polling and elections.

Privacy and ethics


Data collection and storage is an unavoidable part of Petit Poll, but we limit and mitigate this as much as possible. We retain only what is essential, and separate individual level data (such as email addresses) from survey responses.

Specifically, we use and store your email address to ask you to take our surveys, and to let you know the results. If you would like us to delete your email address then simply unsubscribe from your next Petit Poll email, or get in touch and ask. Your email address is not associated with your survey responses - our email list is maintained by a separate service to our survey responses. We store the state that you vote in along with your email address so that you only get emails that are relevant to that state’s election.

Our surveys ask for demographic information at a level that is designed to protect the indentification of individuals. For instance, we do not ask for your age, we ask for your age-group. Again, your email address is not associated with your survey responses. If you have any concerns or suggestions for how we could improve, then please get in touch.

Ethics and political disclosers

We are not affiliated with any particular political party. However we will never work with a candidate that uses sexuality, religion, or gender, to discriminate. Of course, this decision is not black and white, but for instance, we would never work for One Nation, or Donald Trump.

Neither Rohan nor Monica are a member of any political party. However in the past Rohan has been a member of, variously, the QLD, NSW, and ACT branches of the Liberal Party. He volunteered on campaigns for various politicians including Bruce Notley-Smith, Malcolm Turnbull, and John Alexander (no relation). He also helped on the campaign for an ALP candidate by handing out how-to-vote cards for Phil Weightman when a friend was his campaign manager and was short-staffed.

Historical results

Monthly results are available to those who take our surveys - anyone is welcome to sign up. However to give you an idea of our result in 2016, our final primary vote forecasts were 42%, 32% and 13% for the Coalition, Labor and Greens. The actual outcome was 42%, 35% and 10%, so our model was close to the Coalition’s eventual vote, a little high for the Greens and a little low for the ALP. By way of comparison, Newspoll was 42%, 35%, 10%; Essential was 43%, 35%, 12%; and ReachTEL was 43%, 35%, 11%. However, the seat count, not the primary vote, determines who forms government and here we were a little off. Our estimate was that the Coalition would win 80-85 seats. They ended up with 76. Three Tasmanian seats unexpectedly changed hands, as did a couple in Queensland and one in South Australia. Our forecasts could have been better, but they were comparable to those of the established pollsters. We only spent $172 over the course of eight weeks, while they would have charged tens of thousands of dollars for their estimates.