Australian elections – are polls bad, or does media misuse them?

Scott Morrison and his National Coalition winning the Australian election is being reported as a shock, in part due to polls predicting a loss.  Are polls a waste of time? Or does media put too much weight on imprecise indications of how people might vote?

I keep saying that at best polls are an approximate indication of how people may vote in the past, and can in no way predict accurate election results in the future. Polls have well known statistical margins of error, but media reporting on them seem to largely ignore this.

Perhaps more accurate ways could be found to predict election results, but I think that a media obsession with trying to predict what will happen in the future is aas much a problem as polling methods.

RNZ – Australia election: Why polls got it so wrong

It was predicted to be the federal election Labor simply couldn’t lose, but after last night’s surprise Coalition win, the opinion poll may struggle to stand the test of time.

Experts say cost cutting and technological change in the polling process is leading to many inaccurate and misleading suggestions.

Nearly all polls predicted Labor leader Bill Shorten would have an easy win with a 51:49 lead over Prime Minister Scott Morrison on a two-party preferred basis.

I dispute that. Polls generally ask something like ‘if an election was held today who would you vote for?’ – perhaps some polls ask ‘who will you vote for on election day?’ but i have never seen that.

And a 2% gap is well within margins of error, which are usually around 3-4%.

51% with a 3% margin of error means there’s a 95% chance of the result being between 48% and 54%.

49% with a 3% margin of error means there’s a 95% chance of the result being between 46% and 52%.

So there is quite an overlap.

In fact, for three years the polls had picked the Opposition to take government.

Again I dispute that. Over the last three years polls tried to measure who people might vote for in the week or two prior to the poll being published.

They are usually whole country polls. Elections in non-MMP countries like Australia and the USA can be decided in just a few key swing electorates or swing states. \being swing electorates they have a history of impressionable swing voters.

Election campaigns are carefully planned to try to change crucial votes right up until election day. Polls are not designed to examine how people mat change their mind at the last minute.

I obviously have ideas about who to vote for in the weeks and days before an election, but I don’t decide for sure until I vote. There must be others who do similar. Polls can’t get inside our heads.

So why exactly were the polls, as ABC political editor Andrew Probyn put it last night, such a “shambles”?

Former Newspoll boss Martin O’Shannessy blamed the flawed forecasting on the fact that many people’s telephone habits have changed.

“Not everybody has a landline and the numbers that are published are incomplete.”

That might be a part of the problem – but that doesn’t address the ‘trying to predict the future’  misrepresentation of polls.

Polls can only be approximate.

I think that media trying to use polls as precise predictors of future voting is the biggest problem here.

Previous Post
Leave a comment

14 Comments

  1. David

     /  20th May 2019

    The results were pretty close and the polls were too, seems more like a hissy fit from journalists whose team didnt win and looking for an outlet for their frustration.
    I see Sportsbet lost 3 million on the election so the bookies definitely got it wrong and badly.

    Reply
    • Ray

       /  20th May 2019

      Saw a tweet yesterday that suggested maybe changing leaders on the basis of polls might not a good idea.
      Unfortunately polls are used as a drip, drip method of undermining political leaders that the media here and Australia love to use.
      It definitely might be an idea for all political leader to have Bolger’s “Bugger the polls” embroidered on their pillows!

      Reply
    • Blazer

       /  20th May 2019

      don’t ever believe what bookies say.

      They have a huge marketing budget,and the ‘largesse’ is all allowed for.

      Reply
  2. David

     /  20th May 2019

    It also shines a light on the problem with centralized media where they all live in one place, all went to the same universities, all hang out with themselves, follow each other on twitter etc and seldom interact with people from other backgrounds and industries.
    Confirmation bias which in the US literally had people in newsrooms in tears when Hilary didnt win in shock and after Brexit a similar result.

    Reply
    • Alan Wilkinson

       /  20th May 2019

      As Duker linked yesterday there was more than media misuse going on because the polls were unstatistically consistent. The sampling or the processing had been manipulated.

      Reply
      • I hadn’t taken any notice of Australian polls, but there does seem to be an unusual consistency.

        Details: https://en.wikipedia.org/wiki/Opinion_polling_for_the_2019_Australian_federal_election

        It isn’t all media, Roy Morgan is in there too. But perhaps pollsters are trying too much to be seen as accurate predicters after the event. If so that defeats the purpose of polling.

        Reply
        • Duker

           /  20th May 2019

          Your idea that the polls are only ‘predicting’ the time they are taken isnt correct. Because polls were released literally the day before the election and taken over small window of time ( not like NZ where they run the questions for up to 2 weeks)
          Those polls too got it wrong.
          I anything the older polls -over time -did see Morrison closing the gap, which with hindsight was correct.
          Either way PG , your analysis on why they are wrong is flawed

          Reply
  3. Alan Wilkinson

     /  20th May 2019

    NZ Lefty journalists are feeling the hurt too. This pretence to a business reporter is whimpering pitifully, thankfully not paywalled so all can enjoy:
    https://www.nzherald.co.nz/business/news/article.cfm?c_id=3&objectid=12232337

    Reply
  4. Patzcuaro

     /  20th May 2019

    https://en.wikipedia.org/wiki/Opinion_polling_for_the_2019_Australian_federal_election

    This poll of polls has the Labour party narrowly ahead in the two party preferred vote but trending towards the Liberals and in the primary vote the Liberals are ahead and trending higher. It is always easy to see in hindsight but it looks to me as the polls weren’t far out with a narrow lead to Labour but a trend towards the Liberals.

    Reply
    • Duker

       /  20th May 2019

      No . The polls were wrong. When you look at polls by state, they all over represented labour . ‘Within margin of error’ reasoning would require them to have got the Coalition vote ‘too high’ in a few states. Didnt happen.

      As expected Farrar isnt making any stories about ‘how the polls were wrong’ . Instead hes talking up the seats .
      he gets that wrong as well
      ‘Boothby – Libs ahead by 1,311 votes with 74.6% counted
      Chisholm – Labor ahead by 325 votes with 71.1% counted
      Lilley – Labor ahead by 1,323 votes with 74.3% counted
      Macquarie – Labor ahead by 620 votes with 82.1% counted
      Wentworth- Libs ahead by 1,053 votes with 73.0% counted

      Preferences havent been distributed so you cant tell who ahead – those are guesses

      looking at one seat Chisholm and the REAL results
      https://tallyroom.aec.gov.au/HouseDivisionPage-24310-205.htm

      Liberal is 43.25%
      labour is 35.2%
      Green 11.67%
      Which is clearly ‘not ahead’. Yes Green preferences will mostly but not all flow the labour

      Another thing to note is that the numbers of votes as % that are counted isnt that low. When results are declared and all votes counted it will still be ‘less than 100%’

      Reply
      • Duker

         /  20th May 2019

        The main reason around 5% of ‘votes arent counted’ is they are classed as informal as they dont mark all boxes with a number as is compulsory
        The Seat of Chisholm has 4.21% AS INFORMAL and are ‘not counted’

        Reply
      • Patzcuaro

         /  20th May 2019

        You are always going to get local anomalies, Tony Abbott for instance, but when I look at the poll of polls graphs, the election result isn’t surprising, the momentum or trend was with the Liberals going into polling day.

        Reply
        • Duker

           /  20th May 2019

          The person replacing Abbott says she would support the coalition- except climate change issues. Thats is a small L liberal.
          Quite a few of the independents are like that , I think there is only one in Tassie thats labour leaning
          The reality was Abbott was an conservative in a social- liberal seat. The marriage referendum results should have been a warning to him as he was on wrong side

          Reply

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s