Sunday, May 19, 2019

Election surveys go wrong in Australia

Once again, pollsters have gone wrong, this time in Australia.

Surveys conducted mainly by Newspoll, YouGov/Galaxy, Ipsos, Essential, and Roy Morgan over the past three years, consistently (except on a few occasions) put the Labor Party ahead of the Coalition (Liberal-National Party).

But the results are now just the opposite, with the LNP winning a third time in a row.

This Wikipedia page gives you a good idea of what the opinion polls projected over the past few years, and how it all ended up today.

The result has been aptly summed up by the winner, incumbent Prime Minister Scott Morrison, who said, "I have always believed in miracles."

Not the first time pollsters have gone wrong, especially in the recent past. The Brexit referendum in June 2016, American presidential elections in November 2016, are the most famous ones.


Surveys, in general, are a tricky game. It's actually a very specialised statistical science. It not just about talking to some people and coming to a conclusion about how the wind is blowing.

The fact that very reputed polling agencies like Gallup in the United States have got it wrong on a few occasions points to how error-prone the whole operation is.

One reason poll results tend to go wrong is that the sample selected is not a "representative" one. One common misconception is that the larger the sample, the more accurate the projection is.

It's not the size, but how representative is the sample determines the accuracy. Of course, when it is representative sample size is proportionately big, the errors get cancelled out providing a more precise result.

Then, of course, there are always possibilities of people saying one thing to the pollster and voting the opposite way. Either because of a perceived fear factor or a desire to sound politically correct, people might not be truthful about sharing their choice with the surveyor.


Today, India votes in the last of the seven phases. Within a few hours, TV news channels will go all out with the results of the exit polls.

In India, publishing of results of opinion polls or exit polls is banned while the election process is on.

The popular perception is that the ruling coalition led by the right-wing BJP will be back in power. But no one is sure of how big or small majority they would get.


SG said...

Interesting topic. My personal view is people say one thing to the pollster and vote the opposite way.

Hilary Melton-Butcher said...

Hi Pradeep - thanks for posting this ... I'll be back to read properly. Sadly I think ours is a foregone conclusion - albeit it's for the European Parliament ... it'll add another layer into the political morass we're in ... cheers Hilary

KParthasarathi said...

The percentage of error in exit polls is subject to many factors- the neutrality or otherwise of the pollster conducting the survey, the true representative character of the sample, the truthfulness of the voter in revealing his/her choice in what is a secret ballot in a charged atmosphere outside the polling booth with party agents and their men milling around. It is at best an exercise to know the broad trend and nothing more.

Liz A. said...

And some of the problem is not secure elections with tampering happening via cheating or foreign government influence. That's a huge issue right now.

Pradeep Nair said...

- Hi Rajan: Yes, one can't be sure of the exit polls.

- Hi Hilary: Indeed, politics in Europe is in such a mess. I just hope it all will settle down soon.

- Hi KP: Absolutely. Full agree. It's very unpredictable.

- Hi Liz: Yes, technology has another dimension. While it has helped in some ways, it's also wreaking havoc.

rudraprayaga said...

Yes, it can go wrong provided the winning margin is too scanty because the wave can move this side or that side.If the difference goes high probably there is a chance of the poll to be correct.Anyway apt topic at apt time.

Pradeep Nair said...

Hi Sarala: That's a good point.