None of the polls predicted the winner of the 2019 Australian election, the first such failure since 1993 when all the polls started reporting a two-party preferred (2PP) vote estimate of the vote share as well as the parties' first preferences. But the idea that the polls had enjoyed a very good run until 2019 is misleading: from 1993 to 2016, a fifth had predicted the wrong winner. This paper examines the performance of the polls against several measures: the outcome; margins of error; size of the errors; and estimates of the gap between the Liberal-National Party (LNP) and Labor. It shows that about a third of the estimates of the 2PP vote, Labor's first preferences, and the LNP's first preferences, involved errors greater than those attributable to sampling error.
- Australian politics
- opinion polls