Saturday, February 22, 2020

Aggregated poll update

It has been a couple of months since I last did an aggregated poll update. So it is time for a quick update. [Edit: charts updated on 24 February to include the latest Newspoll].

If we start with the two-party preferred (TPP) polls: six months after the previous election we are in the remarkable position of only having poll estimates from Newspoll. No other polling firm has published a national TPP estimate since the polling failure of the 2019 Federal Election. So there is nothing to aggregate.

We can, however, apply the model assumption that population voting intentions are slow to change over time. This will smooth the noise from the Newspoll TPP series. The result is a slow decline in the estimated TPP vote share for the Coalition since the last election, which accelerated over December/January.

On the primary vote side, the Newspoll series has been augmented by a single ANUpoll, from the Centre for Social Research and Methods at the Australian National University. Reading the ANUpoll report closely, it looks there was an earlier poll in October 2019 as well, but I have not seen the details of that poll.

For the purposes of aggregation, I have assumed that the two polls are unbiased on average. That is to say, their house effects sum to zero. This is a big assumption. Furthermore, while we can technically aggregate the individual polls from ANUpoll and Newspoll, with only one poll from ANUpoll, I would urge caution. It is better to focus on the trends (up, down or stationary), rather than the actual numbers. In that vein, the Coalition's primary vote estimate is down. All of the other parties/groups are up since the 2019 election.

The house effect charts show the difference between Newspoll and ANUpoll. Although it needs to be repeated that this result is based on just a single ANUpoll (which may or may not be an outlier). The ANUpoll was less favourable to the Coalition and Other parties, and more favourable to Labor and the Greens. Note: a house effect of -1.5 and 1.5 from the two firms for the Greens means these polling firms are on average three percentage points apart.

Finally, we turn to the attitudinal polling. Both Newspoll and Essential have published regular attitudinal estimates. In respect of the preferred Prime Minister, the key trend has been Scott Morrison's decline.

On the satisfaction side, we have also seen a decline in the satisfaction with the Prime Minister. The Prime Minister now has a net-negative satisfaction score.

I have used the polling data from Wikipedia to prepare these charts.