The Wake-Up Call

Three polls have shown a shift towards the Coalition but our forecast has hardly changed. What's going on?

Over the last week we’ve seen three polls come in from three different pollsters: Roy Morgan, Galaxy, and the closely watched Newspoll. Both Roy Morgan and Newspoll have the ALP two-party preferred at 51% and the Coalition 49%. Galaxy has the ALP at 52% and the Coalition at 48%.

All three pollsters have shown a reduced margin between the two parties compared to their previous release.

Together this gives consistent evidence of a tightening contest between the ALP and the Coalition for the upcoming election.

At the same time, our forecast has hardly moved an inch. What’s going on?

The short answer is that these new polls have been in line with how our forecast interpreted previous polls.

Here’s the slightly longer answer. Our forecast adjusts polls based on the historical accuracy of polls by the same polling firm taken a similar number of days out from an election. For example, the election is now 20 days away so the latest Newspoll is adjusted based on the accuracy of Newspolls around 20 days out from a previous election. Similar adjustments are made to all polls based on the historical accuracy of the polling firm.

Polling three weeks out from an election has usually been pretty accurate compared with the final result. Past polls at this point in the campaign are also evenly balanced between overestimating the ALP vote and overestimating the Coalition vote. This means we’ll be making smaller adjustments to polls from this point on.

It’s not the same for polls further out from an election. Historically, polls further out from an election *tend* to overestimate the ALP’s two-party preferred vote compared to the actual election result. This means older polls included in our forecast were adjusted so they did not favour the ALP as much as they reported.

Because of these adjustments, the evidence inside our forecast has been consistent since the budget. We expect the ALP to win the two-party preferred vote by around 51% to the Coalition’s 49%. We had a similar expectation when polls were reporting the ALP at 52% because the election was still a while away.

Simply put, our forecast anticipated a shift towards the Coalition and that’s exactly what has happened. The polls are matching our expectations.

Self five

We’re going to do a deep dive into how we combine polls to estimate the two-party preferred vote. This will include an examination of the size and spread of polling errors leading up to an election. We might even speculate a bit about why there are predictable shifts in the run up to an election. Keep your eyes out for this at the end of the week.