Why the polls keep getting it wrong
Pollsters should alter their strategies and correct biases within sampling to avoid a repeat of their 2019 election failure, an independent report has found.
The Association of Market and Social Research Organisations (AMSRO) has released the findings of its inquiry into pollsters' inability to predict Scott Morrison's shock 2019 victory.
The Coalition retained power with 51.5 per cent of the two-party preferred vote last May, despite all major polling tipping Labor for a dominant victory.
AMSRO said the error was so uniform it could be considered a "polling failure", but described it as an "anomaly" given the previous precision of Australian pollsters.
It gave 10 recommendations to avoid a repeat, including establishing a code of conduct for polling before the next federal election and publicly available disclosure standards.
The inquiry found pollsters over-estimated Labor's support because they failed to correct a bias towards politically engaged and better educated voters.
In fact, it concluded polling had overemphasised support for Labor over two-thirds of the time since 2010.
Inquiry chair Darren Pennay warned pollsters "should seek to better understand the biases in the sample and develop better … strategies to improve representativeness".
He ruled out a late swing in voter preferences, the impact of "shy conservatives", or people deliberately misleading pollsters as major contributing factors to the 2019 failure.
Similar post-mortems into polling have been conducted in the UK and US.
But AMSRO said its inquiry was hampered by a lack of information from pollsters, which "came as a shock to our international advisers … (and) materially affected our ability to identify the specific factors that contributed to inaccuracy in the polls".
AMSRO president Craig Young at the time the inquiry was launched said: "We hope that this report ultimately provides the polling companies with useful information … to assist them in their efforts to improve the accuracy - and public perceptions of the accuracy - of election and political polling."
Originally published as Why the polls keep getting it wrong