Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Aug 3, 2019 16:14:11 GMT
It has to be pointed out that YouGov were the most accurate pollster in 2014 euros...
Its important not conflate false recall with inaccuracy. After all pollsters got 2017 wrong specifically because they weighted by turnout in 2015
|
|
|
Post by justin124 on Aug 3, 2019 16:28:59 GMT
Ipsos is down to lack of prompting for the Brexit Party, so the Conservative and Brexit Party numbers would basically be the same for every pollster if they prompted. Mori did prompt for BXP in their latest poll, as said in the latest article here ukpollingreport.co.uk/The argument is that Yougov know the past vote of panel members recorded at the time of the General election and use that, whereas the others use the currently recalled past vote. A significant number(7% absolute) have forgotten they voted Labour in 2017, and they get downweighted in non Yougov samples. Tory voters remember how they voted much better. Mori do not weight for past vote at all. False recall is a well known thing and may be affecting Brexit polling as well, but this has not been tested by Yougov as yet as far as i can see. So, just to summarise the conclusion: Survation, Comres, Opinium and maybe some others are overestimating the Labour vote currently. Yougov and Mori are more accurate on the Labour vote, as they were in the Euro elections
Yougov was the only pollster to underestimate Labour in the Euro election.
|
|
|
Post by archaeologist on Aug 3, 2019 18:09:47 GMT
Yes YouGov did underestimate Labour in the Euro Election - by a whopping 1%. Apart from MORI all the others overestimated Labour by at least 3% and at most 11% (ComRes by 10%).
|
|
|
Post by justin124 on Aug 3, 2019 19:22:13 GMT
Yes YouGov did underestimate Labour in the Euro Election - by a whopping 1%. Apart from MORI all the others overestimated Labour by at least 3% and at most 11% (ComRes by 10%). Yougov's final poll covered 19th - 21st May and ended 48 hours before Polling Day. Labour's support was clearly slipping in the final few days with very little effort being mounted by the two main parties. It is not unreasonable to believe,therefore, that a few days before the EU election Labour's rating was 16%/17% - ie when Yougov was still polling. It is also true that Yougov had the Brexit Party at 37% - much higher than the 31.6% outcome..
|
|
|
Post by andrew111 on Aug 3, 2019 21:09:13 GMT
It has to be pointed out that YouGov were the most accurate pollster in 2014 euros... Its important not conflate false recall with inaccuracy. After all pollsters got 2017 wrong specifically because they weighted by turnout in 2015 If you don't take account of false recall correctly then your results will be inaccurate. The same arises if you don't take account of differential turnout correctly. The problem is that turnout and false recall are variables dependent to an unknown amount on other factors. Why is it that in YouGovs data Labour voters are much more likely to recall their 2017 vote than Tory voters? Given current circumstances i would have thought many voters would want to erase voting for either of them from their memory.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Aug 3, 2019 21:25:28 GMT
It has to be pointed out that YouGov were the most accurate pollster in 2014 euros... Its important not conflate false recall with inaccuracy. After all pollsters got 2017 wrong specifically because they weighted by turnout in 2015 If you don't take account of false recall correctly then your results will be inaccurate. The same arises if you don't take account of differential turnout correctly. The problem is that turnout and false recall are variables dependent to an unknown amount on other factors. Why is it that in YouGovs data Labour voters are much more likely to recall their 2017 vote than Tory voters? Given current circumstances i would have thought many voters would want to erase voting for either of them from their memory. this isn't true. Take Survation in 2017. They didn't get the election right because their weighting was right. Their weighting was horribly wrong. More so than Ipsos and their headline figure was almost the worst of the bunch. Solving false recall doesn't mean an accurate headline figure. This is the problem pollsters face. The pollsters that got the last election wrong had weighted more accurately than pollsters that got the result right
|
|
|
Post by andrew111 on Aug 4, 2019 6:09:09 GMT
If you don't take account of false recall correctly then your results will be inaccurate. The same arises if you don't take account of differential turnout correctly. The problem is that turnout and false recall are variables dependent to an unknown amount on other factors. Why is it that in YouGovs data Labour voters are much more likely to recall their 2017 vote than Tory voters? Given current circumstances i would have thought many voters would want to erase voting for either of them from their memory. this isn't true. Take Survation in 2017. They didn't get the election right because their weighting was right. Their weighting was horribly wrong. More so than Ipsos and their headline figure was almost the worst of the bunch. Solving false recall doesn't mean an accurate headline figure. This is the problem pollsters face. The pollsters that got the last election wrong had weighted more accurately than pollsters that got the result right To get accurate results consistently you have to solve all the issues correctly. Sometimes you might make compensating errors that accidentally give a good result. If a pollster gets a particular election wrong outside MOE, that shows their methodology was wrong for that election. If they get it right, that does not necessarily show their methodology was right. It could be a stopped clock which is accurate twice a day.. By not weighting for past vote, Mori avoid a source of potential systematic error, but increase the random margin of error because they have less control on the sample (imo). This particularly affects smaller Parties where the sample is small. By using their large panel, YouGov also avoid systematic errors from false recall, but the going on the YouGov panel attracts people who like answering questions about soap powder etc on a regular basis, and may not be representative of the voting public..
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Aug 4, 2019 7:45:42 GMT
this isn't true. Take Survation in 2017. They didn't get the election right because their weighting was right. Their weighting was horribly wrong. More so than Ipsos and their headline figure was almost the worst of the bunch. Solving false recall doesn't mean an accurate headline figure. This is the problem pollsters face. The pollsters that got the last election wrong had weighted more accurately than pollsters that got the result right To get accurate results consistently you have to solve all the issues correctly. Sometimes you might make compensating errors that accidentally give a good result. If a pollster gets a particular election wrong outside MOE, that shows their methodology was wrong for that election. If they get it right, that does not necessarily show their methodology was right. It could be a stopped clock which is accurate twice a day.. By not weighting for past vote, Mori avoid a source of potential systematic error, but increase the random margin of error because they have less control on the sample (imo). This particularly affects smaller Parties where the sample is small. By using their large panel, YouGov also avoid systematic errors from false recall, but the going on the YouGov panel attracts people who like answering questions about soap powder etc on a regular basis, and may not be representative of the voting public.. in an ideal world yes but in reality that's not possible as we saw in 2017 where weighting produced more accurate sample of who voted but not how they votd
|
|
|
Post by andrew111 on Aug 4, 2019 8:20:51 GMT
To get accurate results consistently you have to solve all the issues correctly. Sometimes you might make compensating errors that accidentally give a good result. If a pollster gets a particular election wrong outside MOE, that shows their methodology was wrong for that election. If they get it right, that does not necessarily show their methodology was right. It could be a stopped clock which is accurate twice a day.. By not weighting for past vote, Mori avoid a source of potential systematic error, but increase the random margin of error because they have less control on the sample (imo). This particularly affects smaller Parties where the sample is small. By using their large panel, YouGov also avoid systematic errors from false recall, but the going on the YouGov panel attracts people who like answering questions about soap powder etc on a regular basis, and may not be representative of the voting public.. in an ideal world yes but in reality that's not possible as we saw in 2017 where weighting produced more accurate sample of who voted but not how they votd I don't think we disagree. Polls are more useful for showing the direction of change than absolute values of support. All the polls showed that Labour were going up fast during the 2017 GE, and they all showed that BXP and the Lib Dems were going up during the Euro campaign this year. They now show the Tories going up (bearing in mind their 5th place in the Euros) at the expense of BXP. Other changes such as LD and/or Green to Lab may be happening but are not yet established.
|
|
|
Post by tonygreaves on Aug 4, 2019 16:02:11 GMT
When I talked about trends before the Euro elections I was roundly denounced by certainly persons here! But I was (I that instance anyway) right.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Aug 4, 2019 18:40:26 GMT
When I talked about trends before the Euro elections I was roundly denounced by certainly persons here! But I was (I that instance anyway) right. who was this certain person
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Aug 12, 2019 21:05:41 GMT
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Aug 12, 2019 21:13:23 GMT
|
|
|
Post by London Lad on Aug 12, 2019 21:53:25 GMT
Q7 response interesting.
"Boris needs to deliver Brexit by any means, including suspending parliament if necessary, in order to prevent MPs from stopping it"
Agree - 44% Disagree - 37%
|
|
|
Post by carolus on Aug 13, 2019 8:03:01 GMT
Q7 response interesting. "Boris needs to deliver Brexit by any means, including suspending parliament if necessary, in order to prevent MPs from stopping it" Agree - 44% Disagree - 37% What a garbage question. This is supposed to be a serious polling organisation, and they're calling the PM "Boris" in an actual real question. Plus, leading (and barely coherent).
|
|
|
Post by Davıd Boothroyd on Aug 13, 2019 8:10:16 GMT
Polling experts on twitter have been taking apart the Telegraph's reporting on this. The question is clearly biased and suggesting an answer; it's an 'agree/disagree' question which always gives misleading answers (see: ukpollingreport.co.uk/blog/archives/10054), and the Telegraph has quietly discounted the 'Don't Knows'. Surprisingly the British Polling Council's rules don't extend to preventing poll sponsors asking biased questions, though the more reputable companies would insist on rephrasing them. Not old Comedy Results.
|
|
|
Post by carolus on Aug 13, 2019 8:18:37 GMT
What a garbage question. This is supposed to be a serious polling organisation, and they're calling the PM "Boris" in an actual real question. Plus, leading (and barely coherent). Well you have to diss it when it provides the 'wrong' response don't you? No, I have to diss it when the question is ridiculous. Can you honestly tell me you think that referring to him as "Boris" is a reasonable thing to do? Would they have referred to "Theresa" or "Gordon" in this way? Can you honestly tell me you don't think the question is a) ambiguous and b) leading?
|
|
|
Post by carolus on Aug 13, 2019 8:19:29 GMT
Polling experts on twitter have been taking apart the Telegraph's reporting on this. The question is clearly biased and suggesting an answer; it's an 'agree/disagree' question which always gives misleading answers (see: ukpollingreport.co.uk/blog/archives/10054), and the Telegraph has quietly discounted the 'Don't Knows'. Surprisingly the British Polling Council's rules don't extend to preventing poll sponsors asking biased questions, though the more reputable companies would insist on rephrasing them. Not old Comedy Results. Notice that the Telegraph article has edited the question, and reported it as "The Prime Minister", not "Boris".
|
|
sdoerr
Conservative
Posts: 148
|
Post by sdoerr on Aug 13, 2019 20:17:45 GMT
Polling experts on twitter have been taking apart the Telegraph's reporting on this. The question is clearly biased and suggesting an answer; it's an 'agree/disagree' question which always gives misleading answers (see: ukpollingreport.co.uk/blog/archives/10054), and the Telegraph has quietly discounted the 'Don't Knows'. Surprisingly the British Polling Council's rules don't extend to preventing poll sponsors asking biased questions, though the more reputable companies would insist on rephrasing them. Not old Comedy Results. How would you rephrase it?
|
|
|
Post by Davıd Boothroyd on Aug 13, 2019 20:34:37 GMT
Polling experts on twitter have been taking apart the Telegraph's reporting on this. The question is clearly biased and suggesting an answer; it's an 'agree/disagree' question which always gives misleading answers (see: ukpollingreport.co.uk/blog/archives/10054), and the Telegraph has quietly discounted the 'Don't Knows'. Surprisingly the British Polling Council's rules don't extend to preventing poll sponsors asking biased questions, though the more reputable companies would insist on rephrasing them. Not old Comedy Results. How would you rephrase it? Original question: "Do you agree or disagree with the following statements? Boris needs to deliver Brexit by any means, including suspending parliament if necessary, in order to prevent MPs from stopping it" How about: Should Parliament reject a revised withdrawal agreement with the EU, which of the following comes closest to your view? • Parliament should decide what to do next, even if it means further delay before Britain leaves the EU * Parliament should be 'prorogued' so it is unable to meet, even if it means Britain leaving the EU without any agreement * Don't know
|
|