|
|
Post by rcronald on Jun 18, 2024 18:10:47 GMT
Their MRP has Labour only winning by 3% in Chingford…
And Labour winning by 2% in Hornchurch…
|
|
|
|
Post by andrewp on Jun 18, 2024 18:13:03 GMT
I mean if the Lib Dem’s win Tiverton & Minehead by more than they win Taunton & Wellington, and the Lib Dem’s win West Dorset by 15% whilst the Tories hold Glastonbury and Somerton, I’ll eat a hat shop.
|
|
|
|
Post by Forfarshire Conservative on Jun 18, 2024 18:13:30 GMT
Some of the estimates are bizarre. Reform at 59% in Ashfield? |
|
edgbaston
Labour
Posts: 4,425
Member is Online
|
Post by edgbaston on Jun 18, 2024 18:23:47 GMT
Maybe we need a ‘common sense MRP’ where there is manual correction on a list of special seats (or even separate constituency polls in these seats as part of the model) |
|
CatholicLeft
Labour
2032 posts until I was "accidentally" deleted.
Posts: 6,732
|
Post by CatholicLeft on Jun 18, 2024 18:30:40 GMT
Some of the estimates are bizarre. Reform at 59% in Ashfield? That on is absolutely barking mad. The trouble with MRPs is that the methodology takes over from common sense when it throws out such divergent predictions without correction. |
|
|
|
Post by John Chanin on Jun 18, 2024 19:33:38 GMT
I on the other hand like the fact that polling companies put out their results without any attempt to massage them. We can then all criticise, and the companies can look to improve their methodology for the future.
20,000 is the minimum for an MRP. It's only 30 per constituency. The more factors in your regression the larger the error factors get, so small MRPs use less factors, and therefore miss the target. YouGov I think used 60,000 for theirs. Ideally I'd want 100,000. But it's expensive to do such mega surveys, although less so for YouGov who I understand have hundreds of thousands on their panel.
|
|
|
|
Post by greenchristian on Jun 18, 2024 19:36:22 GMT
Interesting, has the Greens winning 3, but Brighton Pavilion not among them... In an MRP Brighton Pavilion and Bristol Central should be showing similar trends - the two seats have very similar demographics. If we are on course to win Bristol and lose Brighton then it will be because of the kind of local factors that an MRP model can't take into account. |
|
xenon
Forum Regular
Posts: 427
|
Post by xenon on Jun 18, 2024 20:28:52 GMT
Surely this data has been massaged somehow or else how would it randomly predict that only Clacton and Ashfield out of a dozen or so constituencies with very similar demographics are going to have almost double the Reform UK vote share of the others? Similar story with the Greens in North Herefordshire – there must be some sort of "minor party target seat here" override which is boosting the figures.
In related news, Waveney Valley is down as being a Green/Reform UK tossup – when was the last time when all of the main three parties stood in an English constituency but none of them came first or second?
Other than that and a few other oddities it seems vaguely plausible – Scotland in particular seems a lot less mad than YouGov or Electoral Calculus's attempts.
|
|
|
|
Post by aidypiez on Jun 18, 2024 20:30:43 GMT
In related news, Waveney Valley is down as being a Green/Reform UK tossup – when was the last time when all of the main three parties stood in an English constituency but none of them came first or second? Rochdale by election 2024 |
|
xenon
Forum Regular
Posts: 427
|
Post by xenon on Jun 18, 2024 20:32:15 GMT
In related news, Waveney Valley is down as being a Green/Reform UK tossup – when was the last time when all of the main three parties stood in an English constituency but none of them came first or second? Rochdale by election 2024 Well played, should have seen that one coming. In a General Election then? |
|
xenon
Forum Regular
Posts: 427
|
Post by xenon on Jun 18, 2024 21:13:49 GMT
In answer to my original question, I found this in the methodology which I suppose partially explains some of the high minor party shares, but I imagine these still will have needed to be manually boosted since in a 20k sample, there will only be 31 respondents on average in each constituency.
"As the modelling makes use of a national survey, caution should be taken when looking at individual seats. While MRP is good at taking into account the different demographic profiles of each constituency, with relatively few respondents per constituency, it is unlikely to be able to capture the full local context, especially where there are unique political dynamics. In order to aid this, we have prompted with all candidate names, where known, in 20 constituencies, mainly where there are high profile candidates outside the main traditional parties. These were Aberdeenshire North and Moray East, Alloa and Grangemouth, Ashfield, Boston and Skegness, Brighton Pavilion, Bristol Central, Chingford and Woodford Green, Clacton, Clwyd East, Cowdenbeath and Kirkcaldy, Great Yarmouth, Islington North, Leicester East, Lothian East, Na h-Eileanan an Iar, North Herefordshire, North West Leicestershire, Rochdale, Solihull West and Shirley, Waveney Valley."
I must confess I did have to look up what was so special about Solihull West and Shirley as I hadn't realised Julian Knight was standing as an independent.
|
|
|
|
Post by batman on Jun 18, 2024 23:26:58 GMT
I repeat. Get perhaps 3 or 4 of the more sensible people in this forum in a room for about 2 hours, and together we could come up with a far more accurate set of predictions than any of the MRPs with YouGov being a partial exception, arguably. I myself am confident that on my own I could do better than this.
|
|
|
|
Post by manchesterman on Jun 18, 2024 23:55:58 GMT
I repeat. Get perhaps 3 or 4 of the more sensible people in this forum in a room for about 2 hours, and together we could come up with a far more accurate set of predictions than any of the MRPs with YouGov being a partial exception, arguably. I myself am confident that on my own I could do better than this. I have some spare time  |
|
|
|
Post by Penddu on Jun 19, 2024 1:35:50 GMT
But first of all we would need a poll to predict who these sensible people are. Ususl caveats about sample size and regional sub-samples apply....
|
|
peterl
Green
Congratulations President Trump
Posts: 8,473 
|
Post by peterl on Jun 19, 2024 1:45:58 GMT
Regional sub-samples, ugh too much effort. Can't we just produce a few dodgy bar charts and call it a day?
|
|
|
|
Post by islington on Jun 19, 2024 5:29:08 GMT
Maybe we need a ‘common sense MRP’ where there is manual correction on a list of special seats (or even separate constituency polls in these seats as part of the model) Or maybe - just maybe - we could spare ourselves all this febrile speculation and wait patiently for a couple of weeks, until we know for sure. (Sorry. That's a preposterous suggestion. I don't know what came over me. Forget I said it.) |
|
|
|
Post by batman on Jun 19, 2024 6:22:45 GMT
But first of all we would need a poll to predict who these sensible people are. Ususl caveats about sample size and regional sub-samples apply.... if we were to go for me, Pete Whitehead, one of the Lib Dems and Merseymike it would cover most bases 😊 |
|
|
|
Post by iainbhx on Jun 19, 2024 6:30:40 GMT
But first of all we would need a poll to predict who these sensible people are. Ususl caveats about sample size and regional sub-samples apply.... if we were to go for me. Pete Whitehead, one of the Lib Dems and Merseymike it would cover most bases 😊 ntuk1707's knowledge of Scotland is superb. Pete would be absolutely essential. |
|
|
|
Post by batman on Jun 19, 2024 6:41:32 GMT
Yes we need ntuk1707 for Scotland
|
|
|
|
Post by greenrichard on Jun 19, 2024 8:24:51 GMT
I repeat. Get perhaps 3 or 4 of the more sensible people in this forum in a room for about 2 hours, and together we could come up with a far more accurate set of predictions than any of the MRPs with YouGov being a partial exception, arguably. I myself am confident that on my own I could do better than this. I have some spare time Maybe we should select forum members to do some predictions. Can anyone who predicted the first and second places in Rochdale please step forward… |
|
