tag:blogger.com,1999:blog-4052593945054595675.post8961435873530872867..comments2024-03-28T14:16:10.498+11:00Comments on Dr Kevin Bonham: What Might 2PP Voting Intention Have Really Looked Like In The Last Federal Term?Kevin Bonhamhttp://www.blogger.com/profile/06845545257440242894noreply@blogger.comBlogger7125tag:blogger.com,1999:blog-4052593945054595675.post-18939399428609819582019-07-08T11:05:07.930+10:002019-07-08T11:05:07.930+10:00Mark Textor's is one voice we have not heard s...Mark Textor's is one voice we have not heard so far about the causes of the polling failure and his comments would be interesting, especially as he has been quite scathing about public polling in the past. The problem as usual with internal polling is that it is more or less impossible to validate how much an internal pollster really did have it right - and of course the pollster will want everyone to think that they were accurate. Kevin Bonhamhttps://www.blogger.com/profile/06845545257440242894noreply@blogger.comtag:blogger.com,1999:blog-4052593945054595675.post-29875982343532360132019-07-08T02:56:00.874+10:002019-07-08T02:56:00.874+10:00I remember reading, though I couldn't tell you...I remember reading, though I couldn't tell you where, that Crosby Textor actually had the result right (or close to it), and this is why Morrison was confident he could win. If this is so, then someone at CT might well be able to say when the public polls diverged.Unknownhttps://www.blogger.com/profile/02537910486414527809noreply@blogger.comtag:blogger.com,1999:blog-4052593945054595675.post-63347791180796666362019-07-06T18:24:10.994+10:002019-07-06T18:24:10.994+10:00I find myself wondering if there were not one, but...I find myself wondering if there were not one, but many contributors to a cumulative bias which, through herding, then contaminated the thinking of other pollsters. There were any number of landmark events during that term of government, including the one you have used, but also including the S44 sagas and the Barnaby Joice dramas. If, at each of these crisis moments, conventional wisdom suggested a larger drop in voter support than expected, those conducting the polls may have concluded that their sampling was under-reporting pro-labor sentiments and incorporated a small (but cumulative) correction. This practice would be confirmed in pollsters minds by the by-election results, and in particular by the massive swings recorded. If anything, they are likely to have encouraged them to double down on the practices. Then comes the spill, and it's aftermath, the negative impact of which was then overestimated, and the rest is history.<br /><br />An alternative which suggests itself as marginally possible is demographic drift in the case of a number of voters, whose voting intentions subsequently changed accordingly. If those individuals were then erroneously categorized during polls as still being within their former demographic, it could result in incorrect scaling, which then produces an erroneous poll. This would amount to a social trend that was incorrectly captured and processed by the pollsters, just as there is a trend amongst neutrals within the US to bias more liberal because of dislike and disrespect for President Trump, while his base support remains firm. Such a situation in Australian terms could apply.Mikehttps://www.blogger.com/profile/09569372868831612962noreply@blogger.comtag:blogger.com,1999:blog-4052593945054595675.post-91678495581558995512019-06-30T14:10:46.663+10:002019-06-30T14:10:46.663+10:00It's going to be very difficult to know, at an...It's going to be very difficult to know, at any rate.<br /><br />That's probably the worst case - that unreachable voters broke in a way not predictable by their demographic characters. A better case would be that the pollsters weren't scaling for the right things and had the evidence all along if they knew how to use it correctly.Kevin Bonhamhttps://www.blogger.com/profile/06845545257440242894noreply@blogger.comtag:blogger.com,1999:blog-4052593945054595675.post-81001192449111747322019-06-30T12:33:47.383+10:002019-06-30T12:33:47.383+10:00Hmmm. So,supposing a sizeable bloc of voters who t...Hmmm. So,supposing a sizeable bloc of voters who think how they vote is nobody else's business (say, a certain kind of uptight churchgoer) all jumped back to the Libs immediately after ScoMo's ascension, there's no way we'll ever know?Jack Arandahttps://www.blogger.com/profile/06210027164177789357noreply@blogger.comtag:blogger.com,1999:blog-4052593945054595675.post-69160686552059767992019-06-30T12:00:44.088+10:002019-06-30T12:00:44.088+10:00I am not convinced any of the pollsters were herdi...I am not convinced any of the pollsters were herding to each other before the last 3-4 weeks. Jackman and Mansillo pick up signs of 2PP "underdispersal" going back to at least March but I suspect this is mainly caused by Newspoll's tendency to not move about much from poll to poll (which seems to be a form of self-herding, perhaps caused by some kind of stability correction). <br /><br />I think there is some evidence that there was probably not a big change in voting intention in the last few weeks when the string of very clustered results occurred. Had there been a really big change there would have been something unusual in the relationship between voting before the day and voting on the day, but the gap was similar to previous years. So I suspect most of the error existed before the last few weeks. I don't think that it can be a case of the polls being right until 3-4 weeks out, then suddenly someone puts out a wrong value and everyone else herds to it. (Indeed for that to be the case there would have had to be a very dramatic flip in voting intention partway through the campaign, and that's not something that should happen.)<br /><br />However we can't prove that the house effect was, say, 2 points six weeks out. We can only look at a range of plausible values and see what kind of picture they produce. Kevin Bonhamhttps://www.blogger.com/profile/06845545257440242894noreply@blogger.comtag:blogger.com,1999:blog-4052593945054595675.post-59506900825811038942019-06-30T11:14:17.521+10:002019-06-30T11:14:17.521+10:00At least as plausible as any other explanation, Ke...At least as plausible as any other explanation, Kevin, but I'm a bit puzzled by your reference to changing the correction for house effects. How can you see evidence of house effects, and correct for them, at least in the last few months when the pollsters were all desperately herding? Jack Arandahttps://www.blogger.com/profile/06210027164177789357noreply@blogger.com