Printable version of thread

Posted by: vidcapper Jun 25 2016, 06:12 AM

For the second national election in a row, pollsters got the result wrong.

What do you think went wrong this time?

Posted by: JSG Jun 25 2016, 08:32 AM

There were none of the usual exit polls this time though. It really was just a guess asking a few people here and there.

Posted by: Silas Jun 25 2016, 08:51 AM

For the 3rd vote in a row they were spot on about Scotland though.

Posted by: vidcapper Jun 25 2016, 09:25 AM

It's a lot easier to predict a result than the precise figures though.

We don't expect perfection from pollsters, but they should at least be close.

Posted by: Doctor Blind Jun 25 2016, 09:30 AM

The poll of polls had predicted 52 remain, 48 leave. I thought that there would naturally be a little swing to the status quo of around 4% and thus more like 54 remain, 46 leave.

Polls are never going to be 100% correct, but what was slightly surprising was the markets reacting to what they thought was a certain remain vote after the private polling they had commissioned. When you are looking at a close contest then it inevitably does not take a significant amount of error or bias to lead to an incorrect result, so I guess we should not be too surprised that they were wrong.

Posted by: Silas Jun 25 2016, 09:42 AM

They were very accurate about SNP support in the run up to May 2015 and the constituency polling in May 2016 (and they did demonstrate in April and May a trend of the list vote going away from the SNP and into the Greens with the last few polls predicting no overall majority) and then all along they've put Scotland in the remain camp with a vote in the high 50's and low 60's.

Polling north of the border has been much more accurate in terms of precise figures the past 3 years.

Posted by: Suedehead2 Jun 25 2016, 01:24 PM

Thirty years ago, almost all opinion polling was done in face-to-face interviews on the street or at people's homes. The samples were selected to reflect the population in general. Therefore, an interviewer would know what sort of people they needed to find. They might, for example, have been tasked with finding a white man, in his forties and in social class AB.

By the 1990s, that was all changing. First, there was a move towards telephone polling. That had been resisted for a long time because it missed the not insignificant minority of people not on the phone. When that minority became small enough, they could be ignored without skewing the results. Then the pollsters started noticing a problem. They were frequently underestimating Tory support. Further research showed evidence that people were lying about who they had voted for last time. If the government was unpopular, the number of people saying they had voted for them was significantly lower than the number of people who actually did so. In respond, they started to make adjustments to try to allow for that.

The next moves by the pollsters were not entirely of their own choosing. Going out on the streets to try to find a particular type of individual was time consuming and, therefore, expensive. The newspapers who paid for these polls wanted to reduce their cost. That meant making big changes in the way polls were conducted. Face-to-face interviews were gradually abandoned, first in favour of telephone polling and then internet polling. They also changed it so that the sample was selected randomly. They then adjusted the results to try to "correct" flaws in the sample. If their assumptions about how to adjust the figures are wrong, the results will be wrong.

There is another problem with telephone polling. Many people are now reluctant to answer the phone if they don't know who is calling. Companies still using telephone polling don't really have a way of knowing what adjustments - if any - to make for these missed interviews.

Polling for the referendum posed yet another problem. With general election polling, they can use analysis of past election results (including post-election interviews) to help them decide what adjustments to make. There was no sufficient information for the referendum.

In summary, the polls rely far more on what some people might call "guess work" than they used to do. Unless the papers commissioning polls are willing to start paying more for them, this may continue. Of course, those same papers may also decide that the polls are so useless that they will stop commissioning them at all.

Posted by: Common Sense Jun 25 2016, 03:59 PM

Let's hope all the current US election polls are wrong. Most still saying Clinton will win.

Posted by: burbe Jun 25 2016, 05:01 PM

I don't think polls were THAT wrong compared with the GE 2015. The final poll was within the margin of error and Leave was leading for a number of days last week.

Posted by: Jester Jun 25 2016, 05:36 PM

Troll.

Posted by: popchartfreak Jun 25 2016, 07:22 PM

So you're in favour of electing a man who can barely string a coherent sentence together and is clearly an idiot with a very low IQ.

How not surprising. Dumbing down is all the rage....

Posted by: vidcapper Jun 26 2016, 05:34 AM

Then perhaps the problem was that either no-one believed them, or didn't *want* to believe them?

Personally, I was surprised by the Brexit result as anyone - at the start of the campaign I expected a 10% win for Remain. I was so pessimistic that I didn't even bother staying up to watch the results come in!

Posted by: *Ben* Jun 26 2016, 07:47 AM

I don't think they were so wrong. I think the pollsters always say the error margin is around 2-3% and as the polls were so close it was really not easy to say how it could end until the last minute.

I'm not that surprised about the results I must admit. If you think on the simple fact that people are always more involved in something if they have to vote "against" something than to vote "for" something and with such close polls than it was almost clear that the leave parties have a big chance to win.