A problem with polls is also the problem with trolls. For instance, I am an unreliable reporter.
Looks like you jest, but I think it is interesting to note here that a good number of the ones that appear to support Trump are trolls indeed, more than a few in a recent article with interviews mentioned that they wanted to see discord and see shit happen, forgetting that “may you live in interesting times” is not a blessing but a curse.
The problem for Trump is that he is lacking in important endorsements and those low information voters and trolling ones many times, in the end, do not vote in the election.
Jest? Do you really think I could resist trolling a presidential survey, especially one about Trump this far from the election? I play nice here, but you have a better opinion of my RL self than perhaps you should.
Inspired by Wesley’s post and some discussion following here (post #100 on) I am bumping this.
We have the competitive 2016 primary season pretty much under our belts. It seems to me that polling and the aggregators did surprisingly well.
There was of course on the Democratic side in Michigan of course that major fail which seemed to be a result of a systematic error on how to determine likely voters (given that the last competitive race there was the very odd for MI 2008 year). Otherwise pending anyone else doing a more rigorous analysis I am pretty impressed that 538’s polling plus had 11 races they called with 86 to 94% confidence and were indeed right 10 of the 11 times, just as a 90% probability call would be expected to perform.
And Wang’s 3/28 assessment of the Democratic side’s performance in aggregate was also reassuring of the state of the industry.
Still that was in aggregate … Were there any trends in sorts of errors that may be telling for the future? How will polling and the aggregators do in the general where again previous likely voters screens may be not as valid as they previously have been? Can they predict how much increased turnout will occur in previously low turnout demographics (e.g. Trump’s non-college educated White strength)? How many previously likely voters will sit it out?
Any predictions of how good the predictions will be?
A serious question can be asked. What public good does polling do?
Polling has been responsible for much of the horse race journalism that recent elections have devolved into. Many people complain that knowing the outcome in advance, even apparently knowing it, lowers voting and the enthusiasm for the election. How is that a good thing?
More sophisticated polling may have its uses. Candidates should be aware of which issues the public are truly interested in, as opposed to hot button issues. They might also want to know who they are alienating with their positions.
I don’t think the real effects of headlining polls are well understood - do they get lost in the flood of info or stand out as the rare piece of meaningful information?
Overall, though, the best response to the future demise of polling may be good riddance.
Polling is good for campaigns to figure out where to put their resources. It’s good for the public because it gives the public an idea of how their fellow Americans think about issues or candidates, and I think it’s a positive thing to know what other Americans think.
What’s interesting to me, is that, although the reasoning for a growing inaccuracy in polling seems sound, I have yet to see the effects. Generally, the primary season polling has been fairly accurate. Trump and Clinton have mostly won the states that the polling suggested they would, etc.
Aren’t the polls just as accurate as they were before in showing the voting trends? Now the trends are not such simple paths, the electorate has far more information, and misinformation in front of them, every detail of candidates lives are exposed, and the voters don’t march in lockstep to the polls the way they used to. If you only use polls to predict the actual vote counts they’ll certainly look less reliable, but the data may show that turnout and actual votes are not so well settled and that the actual results can’t be so precisely predicted. I know pollsters don’t want to say that for all the effort they put in that the election can’t be called ahead of time, but that’s what the data actually shows isn’t it?
Well, it turns out that trolls and low information voters did help Trump.
But then one has to realize how good the polls then were in pointing at the Trump trend among the Republicans. What it is important to take into account is that back then many did expect the polls to change as early polling many times is not reliable.
But these are not usual times and I do think now that the polarization that is happening is being reflected in polling, the wild swings seen in past elections as in the times of Carter vs Reagan are being replaced by more stable trends like the ones seen in Obama vs Romney.
And that takes us to what IMHO should not be dismissed as the Trump trend among Republicans was; specifically the fact that at the same time that the Trump trend was being noted by the pollsters the very stubborn trend of having Hillary Clinton ahead of Trump was set, and it has been ages when Trump was closer to her by 3 points, nowadays Clinton is regularly 7 or more points ahead of Trump.
I had a theory that in 2004 exit polling influenced turnout. Early exit polls showing Ohio results were released on the internet while TV was still withholding data. When younger voters who were more likely to vote Democratic saw the exit polls they stayed home, while the older voters who watch TV news turned out in their usual numbers.
[QUOTE=;19308268]
… the wild swings seen in past elections as in the times of Carter vs Reagan are being replaced by more stable trends like the ones seen in Obama vs Romney. …
[/QUOTE]
Even back then maybe not so wild as we may think. Again, from Wang’s PEC blog, the graphic being most telling, looking at elections since 1952.
Of note is the volatility in the polls that occurs relatively commonly about 3 to 4 months before the general, before settling back to where they had been more or less.
Okay some of that are things like in 1992 Perot dropping out and coming back in, and some more routine convention bounces. And 1980 Reagan/Carter had the Iranian hostage crisis, rising inflation, and Anderson as externalities.
In any case, I am not so sure about using “Trump” and the word “stable” anywhere near each other … 
Easy. “When Trump goes out to the stable, he gets jealous from seeing how well-hung all the horses are.”
It’s certainly an invaluable tool as far as campaign planning is concerned, which will be enough to ensure its existence. Unfortunately, many of those polls (especially the more trusted ones) are public and thus liable to profit motive.
Second-biggest, anyway.