2016 Postmortem
Related: About this forumNew Bloomberg NH Poll - Sanders 41% Clinton 36% Biden 10%
Bloomberg Politics ?@bpolitics 1h1 hour ago
@BernieSanders leads the pack in the new @bpolitics/@SaintAnselm poll of New Hampshire Dems http://bloom.bg/1GgIWON

Response to kenn3d (Original post)
Name removed Message auto-removed
FiveGoodMen
(20,018 posts)(I hope this one is right!)
JaneyVee
(19,877 posts)riversedge
(74,171 posts)I have never heard of this organization.
...The poll was conducted by Purple Strategies from Oct. 15-18 and surveyed 400 likely Democratic primary voters by landline and cell phone. The margin of error is +/- 4.9 percentage points.
aspirant
(3,533 posts)Now some science is trusty and others aren't
in_cog_ni_to
(41,600 posts)They have no shame.
sufrommich
(22,871 posts)aspirant
(3,533 posts)sufrommich
(22,871 posts)success as other polling orgs.
aspirant
(3,533 posts)not psychics, right?
sufrommich
(22,871 posts)received,some polling orgs are better than others. Silver grades on their historical accuracy and methodology.
aspirant
(3,533 posts)Are scientists and psychics artistic too?
sufrommich
(22,871 posts)I'm not going to play,here's Nate Silver's explanation for his poll ratings:
FiveThirtyEights pollster ratings are calculated by analyzing the historical accuracy of each firms polls along with its methodology. Accuracy scores are adjusted for the type of election polled, a firms sample size, the performance of other polls surveying the same race, and other factors. We also calculate measures of statistical bias in the polls.
http://fivethirtyeight.com/interactives/pollster-ratings/
aspirant
(3,533 posts)"and other factors" What other factors?
in_cog_ni_to
(41,600 posts)arses.
dsc
(52,788 posts)does the poll company do polling for candidates, have they done so for the race or politicians in question would be some. Others would be are they accurate in place A but inaccurate in place B.
aspirant
(3,533 posts)What's different between a scientific poll done for candidates or the media
dsc
(52,788 posts)in order to be able to sample it. New Hampshire allows people to vote in either primary with no restrictions. Also a small but significant proportion of the populace has a history of voting in one primary one time and the other a different time. That means no matter how scientifically a poll is done, it could be reflecting the wrong population and thus wrong. Polls done for candidates tend to be more favorable toward that candidate (by question order, sampling, forcing or not forcing undecideds etc.)
Adrahil
(13,340 posts)I'm not being snarky. Read Silvers "The Signal and the Noise." The basic answer is that part of accurate polling is building an accurate model of the electorate. Subtle differences in such models can significantly affect the outcome. But polls, like many models DO have a control: actual results. If predictions routinely diverge from results, the pollster clearly is not building an accurate model. Or they are not sampling in a way that satisfies the model.
If you want to just reject polling or convince yourself to reject polling because it conflicts with your predetermined desired outcome, I cannot help you with that. Go find some newspaper's self-select "poll" that matches your desired outcome and be happy. Good luck with that.
thesquanderer
(12,473 posts)which means that figures can be wrong. But they can also tell you the odds that they will be wrong. You need to understand things like margin of error and confidence level. (An unscientific poll, by comparison, has no figures for margin of error or confidence level, because the lack of a scientifically valid methodology means the there is no way to calculate accuracy.)
Also, it can be hard to directly compare different polls, because they are not necessarily asking the same exact questions, in the same order, to the same population. For example, "registered voters" is different from "likely voters" (which in turn can be assessed different ways); "registered Democrats" can be different from "people eligible to vote in the Democratic primary" (because many states permit non-Dems to vote in the Dem primary), and so forth.
riversedge
(74,171 posts)Nate Silvers grade of PPP is a B-, which is considerably low for an entity based solely on polling.
Source: http://fivethirtyeight.com/interactives/pollster-ratings/
Also, I find it a little disingenuous when you can say that you have never heard of Bloomberg, Purple Strategies (which is ALWAYS who bloomberg conducts their polls with), or St Anselm.
Further Reading on Anselm: http://www.anselm.edu/News/Bloomberg-Partnership.htm
Poll Results: http://www.scribd.com/doc/286125782/BloombergPolitics-SaintAnselmNewHampshirePoll
riversedge
(74,171 posts)jkbRN
(850 posts)Lol, I beg to differ.
riversedge
(74,171 posts)do not know everything you know you call me disingenuous. You are rude and I am done talking to you. Bye
.....Also, I find it a little disingenuous when you can say that you have never heard of Bloomberg, Purple Strategies (which is ALWAYS who bloomberg conducts their polls with), or St Anselm.
jfern
(5,204 posts)If you average the 4 post debate polls (PPP included), Bernie narrowly leads.
http://elections.huffingtonpost.com/pollster/2016-new-hampshire-presidential-democratic-caucus
Adrahil
(13,340 posts)Clearly there has been a shift since the debate. Hard to say if it was the debate, or the admission by elements of the GOP that the Benghazi committee is politically motivated witch hunt.
thesquanderer
(12,473 posts)WBUR poll has them at 38% HRC to 34% BS with a 4.9 MOE... essentially tied, for the purposes of this poll (i.e. the difference is smaller than the MOE which means there is no actual confidence as to who is ahead)
Bloomberg poll has them at 41% BS to 36% HRC with a 4.9 MOE... again, essentially tied.
The two HRC numbers (38 and 36) are also well within each poll's MOE,
The two BS numbers (34 and 41) are more surprisingly slightly outside of MOE... but remember, even MOE only tells you what the pollsters are (typically) 95% sure of, with therefore an acknowledged 5% chance that the true value is actually outside the MOE, so a true number that is only slightly beyond MOE, while unlikely, is still far from an impossibility.
riversedge
(74,171 posts)Clinton leads the field on leadership questions, with 49 percent of those surveyed saying shes most ready to be president. Biden, despite being a heartbeat away from the presidency for nearly seven years and having served in the Senate for nearly four decades, is seen as most ready by 21 percent of respondents, barely ahead of Sanders, whos at 20 percent. Clinton also has an advantage in being seen as knowing best how to get things done in Washington and in handling Russian President Vladimir Putin. And Democrats see her as their party's strongest potential standardbearer: 50 percent of respondents say that Clinton would be best capable of beating the Republican nominee in the general election. Sanders is next at 19 percent.
Robbins
(5,066 posts)we have had 2 NH polls out today and they are complete opposite of each other.
aspirant
(3,533 posts)These lame phone polls are nothing but propaganda and give their paymasters exactly what they want.
We can openly bribe politicians but pollsters with absolutely no oversight are presented as scientific Saints, ....right.
DCBob
(24,689 posts)So Repubs can vote in Dem primary and vice-versa. In the polling Republicans and Independents can say they are planning to vote in the Dem primary so their choices will get counted in the polling totals. What could be happening is some polls may include Republicans and Independents and some may only include registered Dems. That might account for the big differences. I havent checked the polling details.. just a guess on what might be going on.
mythology
(9,527 posts)Almost every poll is to the 95th percent. Thus statistically 1 in 20 polls gets an inaccurate sampling of the universe of voters.
For example flipping a coin is roughly a 50 50 chance. Some amount of the time you will flip heads 9 times out of 10 times. If you do that, it doesn't change that flipping a coin is roughly speaking a 50 50 proposition. But if the 9 out of 10 heads are your sample, you could could conclude that flipping heads or tails is a 90 10 proposition.
aspirant
(3,533 posts)who oversees and picks out the lame 1-20?
How do you prove the 1-20 statistically if you don't have the names and contact #'s of all respondents insuring an accurate sample?
Agschmid
(28,749 posts)Which I said earlier.
Both sides have work to do.
thesquanderer
(12,473 posts)See post #44
RandySF
(71,707 posts)jfern
(5,204 posts)He was up 8 points with Biden and 10 without Biden.
Elmer S. E. Dump
(5,751 posts)Quixote1818
(30,556 posts)wolfie001
(4,019 posts)".....and they're rounding the first turn....."
workinclasszero
(28,270 posts)
aspirant
(3,533 posts)workinclasszero
(28,270 posts)Just like Bernie fans
kenn3d
(486 posts)Last edited Tue Oct 20, 2015, 10:23 PM - Edit history (1)
The data from any individual poll cannot be directly compared to data reported by other polls/pollsters using differing methodologies. And while Suffolk* had not polled NH since mid-June, the trend from that poll to the October one is Sanders +4, Clinton -4, for a net change in spread (as calculated by this pollster*) of Sanders +8.
Likewise, the prior Bloomberg/Anselm NH poll (Sanders 24 Clinton 56) shows a trend of Sanders +17 Clinton -20 for a net change in spread of Sanders +37.
Of course such comparisons 4 or 5 months apart are not really too useful. But comparisons between different polls/pollsters are not useful at all, and no statistical trends can be drawn from them.
The much better measure is the aggregate or "poll of polls" calculations reported by RealClearPolitics and HuffPollster which plot composite averages from the most recent polls, to produce somewhat more meaningful trendlines and spreads describing the state of any given race. The current spreads reported by these major aggregators are RCP: Sanders +1.2 and HuffPollster: Sanders +8.
So Sanders still leads NH, but Clinton has narrowed that lead significantly since the first debate.
The early state races are really just getting started and will undoubtedly change further as the primaries near, and other states will perhaps change even moreso once the early states have been decided... (by actual voters, not pollsters).
Anything can happen.
99Forever
(14,524 posts)Even those like this one that says what I would prefer.
I don't believe any of them. Every one of them that gets posted is done with an agenda.