Podcast #28: Polling after the federal election

1. What DID go wrong with the polls?

   Ben, Kevin

   I listened to the podcast—very interesting—but got the impression that, while it is agreed that there was SOMETHING wrong, the nature of the “something” is either obscure or jumbled. This morning we have an SMH story that more or less implies that the ALP was told by YouGov that it was in deep trouble; we also have the latest Newspoll.

   These things caused me to rouse myself from my post-election torpor and have a closer look at what I was seeing in my own polling tracking tools. I used the usual suite of tools found in the psephologist’s toolbox (I actually INVENTED some of these as long ago as 1992). For all of these tools, I made regression projections based on trends and displayed the projection with SDs and 95% Confidence Limits. There were some 8 of these (of course, none are independent of the others) and all except 1 showed a lower CL of greater than 50%. The exception was the projection made on the long run trend in ALP TPP in the campaign period (only) of all elections 1992 to 2016. The 95%CLs for this were (48.4% to 52.0%). The 2019 election result was inside that envelope.

   The driving force for me doing such an analysis is the old adage that “the gap always closes”. It was never true arithmetically but, if it is meant to imply “the Coalition always gains at the expense of the ALP” during the election period … then it has been true for 6 out of 8 elections, the average being about 0.4% per polling period.

   This doesn’t mean I wasn’t stunned when I walked out from scrutineering a booth in Warringah, but it does imply I SHOULD NOT have been stunned.

   Geoff Lambert

2. what does the post election polling mean?
   but the pattern for most governments seems to be they are elected then reelected 3rd election is close
   the 4th election is closer and a 5th election win even less likely
   in the case of miracle wins eg 1993 nsw 2007 and aust 2019…………. a bad bad reckoning comes for the winner. 1996, nsw 2011………………

3. I was very surprised at result. All polls indicated an ALP victory, voter behaviour at Bribie pre poll was less overtly pro Liberal than in previous elections and at Morayfield pre poll only sign of change was visible support for One Nation amongst voters. Only sign that ALP was losing was behaviour of Trade Union put Liberal last campaigners whose aggression was an indication of problem. However I did not pick this up till after results were in. Shorten’s Town Hall Meeting In Caboolture was a roaring success. I guess this is living in a bubble but I thought ALP were romping home. The only other time I have been so wrong was when I thought Americans were too intelligent to elect Trump.

   ALP’s review of results is a useful document.
   It is a pity that Mass media who are so keen to have freedom of info and freedom to commit theft for themselves are as secretive as Government when it comes to their own failures.

   If the cause of this failure was herding it means polling organisations do not even believe their results themselves.

   Candidates are notoriously unreliable predictors of election results. They assume that because a voter is civil to them they are voting for them. It is only a small percentage of voters who slam doors in candidates faces.

   ALP is preparing for Qld State election and I get impression that Palasczuk has bitten the bullet and is taking on the Green eco-fascists. She needs to get a real working class candidate in South Brisbane and push Jackie Trad out to Townsville.
   Vince Gair was last working class candidate in South Brisbane and he was a railway clerk.
   ALP needs a shearer or trucker in South Brisbane and Jackie Trad needs to have to Front up to the public bar at the Town and Country.
   First sign of Council Campaigning visible in Strathpine other day. No doubt the seat currently vacant because of court case is attractive.
   Last time this ward had only one candidate.

4. I’ve been offline a lot the last week. Re Goeff’s post, I’ve also noticed this “ALP Fail Factor” (as I call it) in a lot of historic polling analysis – in summary, the Coalition has won a lot more federal elections than Labor, but hasn’t had anything like the same average advantage in polling at any point of pre-election cycles. Labor rarely noticeably outperforms its polling, the Coalition often does.

   A few things have mitigated against using it prominently in my 2019 projections. Firstly it isn’t apparent in state elections. Secondly it wasn’t apparent in the 2013 and 2016 final polls. Thirdly the rapid recent turnover in available pollsters and methods (eg 2019 Newspoll is not 2013 Newspoll) means that it’s been hard to say whether it is something innate in polling of Australian federal elections, or something specific to a set of past pollsters that, possibly excepting Morgan, no longer applied.

   In the previous cycles I was looking at projections with the “fail factor” factored in and they produced worse projections with more data, so it was easy to suspect they were simply overfitted. Unfortunately at this election they would have outperformed projections without them (see eg Peter Ellis’ model, which outperformed those of us who thought house effects weren’t much of a thing anymore, even though it was only using data from the previous four elections.)

5. Ben and Kevin,

   Shock – horror! But, I think I must have one of the very few people who managed to predict the result of the Federal Election.

   Sure Shorten was ahead on social media followers, as was the clear favourite, but one was especially clear and that was (like in the NSW Election), the Coalition simply conducted a far superior campaign then that of the Labor Party.

   It’s as simple as that.

   Elections are won on the 6.00pm news and it was there where Shorten’s lack of temperment was exposed big time – think that question with the channel 10 reporter.

   Also, the Coalition in Scott Morrison sounded far more confident in his campaign.

Comments are closed.