|
Post by Robert Waller on May 4, 2015 19:49:56 GMT
No, we don't know. But I'll say again, when in 1987 I took polls for Harris in Cheltenham, before we put the candidate names on we got the 'wrong' winner, afterwards the 'right' one. We felt we couldn't put names on before nominations closed. Of course, this could have been 'swing', but there was no such movement in national shares over the relevant period. In fact, if I remember correctly, it went the other way a little.
Anyway, I've already backed up this judgment on Hallam in the betting market, so we'll see on Thursday night in this case ...
|
|
YL
Non-Aligned
Either Labour leaning or Lib Dem leaning but not sure which
Posts: 4,209
|
Post by YL on May 4, 2015 20:00:17 GMT
Very interesting. Differences from Ashcroft: General Question: Labour - 34% (-3%) Lib Dem - 32% (+2%) Conservative - 21 (+1%) UKIP - 8 (+-0%) Green - 4% (-1%) Constituency Question: Lib Dem - 42% (+6%) Labour - 35% (-2%) Conservative - 12% (-3%) UKIP - 7% (+-0%) Green - 3% (-1%) So even putting aside the differences in the headline figures, it looks pretty clear that naming the incumbent does make a difference. And let's bear in mind that Clegg is not the most popular incumbent! I think that naming the candidates at this stage may well be the right thing to do, but I'm not convinced there's much evidence for it making a difference. There are other methodological differences between the polls (note that ICM's turnout and don't know adjustments turned a tie into a 7 point Clegg lead; Ashcroft does these things too but it didn't make that sort of difference) and, while we should be cautious about comparing them given the different question order, ICM's general question is also considerably better for Clegg than Ashcroft's. Also ICM's sample size is only 500 (and once you take don't knows, won't votes and won't says into account it's effectively less than that) so it has a pretty big margin of error (and Ashcroft's margin of error isn't that small either).
|
|
|
Post by greenchristian on May 4, 2015 21:38:06 GMT
No, we don't know. But I'll say again, when in 1987 I took polls for Harris in Cheltenham, before we put the candidate names on we got the 'wrong' winner, afterwards the 'right' one. We felt we couldn't put names on before nominations closed. Of course, this could have been 'swing', but there was no such movement in national shares over the relevant period. In fact, if I remember correctly, it went the other way a little. Anyway, I've already backed up this judgment on Hallam in the betting market, so we'll see on Thursday night in this case ... Did you do a second "in your constituency" question in the unnamed version? If not, then this isn't evidence that names make a difference in themselves. If so, then it's certainly evidence (though in evaluating, it's more important whether the result is within the margin of error, rather than whether the right candidate ended up in front).
|
|
|
Post by thirdchill on May 4, 2015 21:48:26 GMT
What an eye opener. I'd love to see more polls showing what effect naming the incumbent has, especially in Conservative and Lib Dem marginals, plus all those seats the SNP are supposed to be sweeping up from Labour. My guess would be an incumbency effect which benefits the parties who hold the seats. Conservatives doing a bit better in their held seats with labour second, Labour doing better in their marginals with the conservatives second, and a general increase in the Lib Dem percentage in their held seats (with significant variations). Of course there will be massive variations, with some MP's not getting an incumbency bonus at all, and possibly a few having a negative effect.
|
|
|
Post by Robert Waller on May 4, 2015 21:49:39 GMT
Green Christian It's a bit hard to remember after 28 years (!), but I think I'd have asked the 'in your constituency' as the only question in the run-up surveys (sounds like me, constituencies, you know). More sure about your second question - before we put names on, the Lib Dems were 5 ahead, after the Conservatives were about 7 ahead. Well outside margins of error, though as I said there could conceivably have been other reasons for the change, though they were only a week or so apart as I recall. Charles Irving (C) won by 7.9% in the end.
|
|
|
Post by greenchristian on May 4, 2015 22:16:52 GMT
Green Christian It's a bit hard to remember after 28 years (!), but I think I'd have asked the 'in your constituency' as the only question in the run-up surveys (sounds like me, constituencies, you know). More sure about your second question - before we put names on, the Lib Dems were 5 ahead, after the Conservatives were about 7 ahead. Well outside margins of error, though as I said there could conceivably have been other reasons for the change, though they were only a week or so apart as I recall. Charles Irving (C) won by 7.9% in the end. That sound like adding the candidates names had the same kind of effect that the second question has in many constituency polls, then. I'm sure we've seen a couple of constituency polls (or possibly multi-constituency polls) in this election that have asked the "in your constituency" question as the only question, and got results that look like the general voting intention on polls that did have a second question. It's possible, (and, in my view, likely) that names would add to the constituency factor. But I would also expect that the second question catches almost all of that effect even without names. I doubt there are many constituencies where it could make more than a couple of points difference.
|
|
|
Post by Robert Waller on May 4, 2015 23:04:36 GMT
I don't entirely understand your point, but I do believe that adding names might not make very much difference in Lab-C marginals. LD seats are a different matter. We shall see how incumbents do compared with the Ashcroft polls in a few days' time anyway. I predict quite a few will end up faring significantly better, and I'll leave it there.
|
|
|
Post by greenchristian on May 4, 2015 23:27:23 GMT
I don't entirely understand your point, but I do believe that adding names might not make very much difference in Lab-C marginals. LD seats are a different matter. We shall see how incumbents do compared with the Ashcroft polls in a few days' time anyway. I predict quite a few will end up faring significantly better, and I'll leave it there. My point is that that we've had constituency/marginal polling this election that asks the "in your constituency" question as the only question, and came back with results that looked far more like normal polling than it did constituency/marginal polling. So if you were only asking the one question back in 1987, it's quite likely that your results without names reflected national preference (even though you were asking specifically about their constituency), whilst your results with names reflected local preference. Both methods (adding names or doing two questions) capture the incumbency/constituency effect. It's possible that adding names will make the constituency effect stronger by a point or two, but it seems unlikely that the difference between a second question with names and a second question without them would be much more than that. Hopefully this clarifies what I was trying to say.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 5, 2015 10:20:37 GMT
|
|
|
Post by Davıd Boothroyd on May 5, 2015 10:52:44 GMT
Don't think we're expecting any more constituency polls though.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 5, 2015 12:02:41 GMT
Don't think we're expecting any more constituency polls though. Couldn't think where to put the tweet.
|
|
carlton43
Non-Aligned
Posts: 48,099
Member is Online
|
Post by carlton43 on May 5, 2015 12:07:12 GMT
Don't think we're expecting any more constituency polls though. Couldn't think where to put the tweet. Oh!! But I can.
|
|
Richard Allen
Banned
Four time loser in VUKPOTY finals
Posts: 19,052
|
Post by Richard Allen on May 5, 2015 12:19:36 GMT
Don't think we're expecting any more constituency polls though. Couldn't think where to put the tweet. Leaving it on twatter would have been the best option.
|
|
|
Post by East Anglian Lefty on May 5, 2015 12:28:44 GMT
I think the key question for pollsters is how late you can release your final poll and still blame any errors on a late swing.
|
|
|
Post by Davıd Boothroyd on May 5, 2015 12:37:29 GMT
According to Smithson there's an Ipsos MORI poll due out on Thursday morning - seems an odd time to choose. It will be caught up in the broadcasting restrictions on polling day.
|
|
|
Post by Andrew_S on May 5, 2015 13:00:59 GMT
According to Smithson there's an Ipsos MORI poll due out on Thursday morning - seems an odd time to choose. It will be caught up in the broadcasting restrictions on polling day. Do the broadcasting restrictions apply to the internet? Because most online forums seem to have featured plenty of speculation and rumour on previous election days IIRC.
|
|
The Bishop
Labour
Down With Factionalism!
Posts: 36,309
Member is Online
|
Post by The Bishop on May 5, 2015 13:09:27 GMT
According to Smithson there's an Ipsos MORI poll due out on Thursday morning - seems an odd time to choose. It will be caught up in the broadcasting restrictions on polling day. They have done that since the 1990s at least, haven't they? Is it also true that YouGov are producing a final VI poll around the close of poll on Thursday??
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 5, 2015 18:54:44 GMT
Yes, Mori are doing their customary attention seeking thing of putting a poll out on election day. Ashcroft is too apparently.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 5, 2015 20:35:10 GMT
So are the media allowed to mention these polls which have been produced. I thought whilst voting happens, it had to be radio silence on polls and the such like.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on May 6, 2015 10:41:00 GMT
|
|