General Discussion
Related: Editorials & Other Articles, Issue Forums, Alliance Forums, Region ForumsAbout that Gallup poll (distribution of Dems vs Traitors)
For results based on the sample of -- 946 -- registered voters, the margin of sampling error is ±4 percentage points at the 95% confidence level.
For results based on the sample of -- 443 -- Democrats and Democratic leaning independents, the margin of sampling error is ±6 percentage points at the 95% confidence level.
For results based on the sample of -- 527 -- Republicans and Republican leaning independents, the margin of sampling error is ±5 percentage points at the 95% confidence level.
No wonder his approval is at a high!
Mike 03
(16,616 posts)unblock
(52,206 posts)simplifying a bit, but basically they take the results for republicans, divide by the number of republican respondents, then multiply by the actual percentage of republicans in the country. then they do the same thing for democrats and combine the results.
having more republican respondents merely reduces the uncertainty of republican responses. it doesn't give them more weight in the final results.
the whole "they polled the wrong people" thing is misleading and needs to stop. thanks for providing some clarity.
Bradical79
(4,490 posts)Gallup's polling is a legit methodology, but is more useful when combined with other sources. At the moment it's an outlier compared to other legit polls. Picking the worst results, or most favorable results too, isn't helpful at all.
RAB910
(3,501 posts)I am betting they rely too heavily on land lines
unblock
(52,206 posts)and then they normalize the results.
much more convenient to solve the problem with math rather than to keep calling until you get exactly the same percentage of dems vs. reps. as in the population at large.
i don't think they're restricted to landlines anymore, but i do think that older people are more accessible and willing to respond to polls.
but again, they normalize for that.
ThoughtCriminal
(14,047 posts)Land-liners that are more likely to answer an unknown number, greedy, and gullible.
LiberalFighter
(50,912 posts)I doubt it. They are probably using percentage they came up with that is 20 years old and outdated.
unblock
(52,206 posts)But nothing close to that stale.
They're pretty good at Gallup when it comes to statistics and methodology.
Where they have problems is in things like question phrasing and order. For instance, if the previous question was about a trump scandal or about the roaring stock market or the coronavirus? Because that previous question might have an impact on the "approval" question.
Not to single out Gallup on this point. It's a general problem that a lot of polling companies get wrong.
Sometimes on purpose...
LiberalFighter
(50,912 posts)That imo is being generous.
And yes, their numbers and other poll companies have stale numbers.
democrattotheend
(11,605 posts)Reputable pollsters don't usually weight on party ID, because party ID tends to fluctuate depending on whom people plan to vote for (or their opinion of the president). Usually they only weight on static characteristics such as gender, race, age, etc.
unblock
(52,206 posts)Well I wouldn't say "reputable" pollsters don't do it. There are pluses and minuses to weighting for party id directly, it's not in and of itself disreputable.
But yes, Gallup doesn't directly do the adjustment I described, that's a simplification.
They achieve much of the same affect by weighting the sorts of factors you list. If they oversampled republicans, most likely the oversampled men or older people, factors which correlate with being Republican. So when they weight for those, most likely they are effectively adjusting for party id in practice
underpants
(182,788 posts)Gallup. Says it all but thanks for the details.
unblock
(52,206 posts)different people take it to mean different things, some people even at different times.
some people approve of the president period, because they figure to do otherwise would be unpatriotic.
some people approve based on performance in the last week or month or so
some people approve based on one issue or quote that stuck in their craw years ago
some people approve based on party alone
some people approve based on whether or not the country is "at war"
some people approve based on how their doing financially
etc.
i get the there's enormous amounts of data on presidential "approval" and it's very well studied, but to me, it's a skyscraper built on quicksand.
0rganism
(23,944 posts)the MOE is going to run higher on things that only apply to the smaller subset rather than the polled sample overall. that's not Gallup-specific.
Bradical79
(4,490 posts)It's gone full stupid a number of times now. Today with the Gallup panic, and I just responded to someone who's, been here awhile, posting "breaking news" about Biden from a small town Kentucky mayor who was citing the Daily Caller (white supremecist Tucker Carlson's site). And a bunch of people were treating it as legit.
I give up lol.
MoonlitKnight
(1,584 posts)And work our asses off and make sure our voters show up and vote.
No complacency this time. None.
Awsi Dooger
(14,565 posts)That's what happens with incumbents during an economy that people are happy with, especially if that incumbent's party has been in power only one term. It is easily the most favorable situational scenario in American politics. The benefit of a doubt is astronomical.
People who continue to obsess over daily variables and listen to Rachel Maddow will continue to have the real world sail comfortably over their head. Those daily developments mean absolutely nothing in the big picture. Issues mean squat. Situational factors dictate one political outcome after another and they can be easily identified years in advance. Democrats are in favorable scenario in 2024 no matter what happens this year. But 2020 always set up as awful, once Barack Obama was re-elected in 2012.
Once Obama was re-elected it meant Democrats would either be seeking a 4th consecutive term in 2020, or the Republican would be in that catbird seat of incumbent whose party has been in power only one term. Basically it's a minor miracle if we win in 2020. It would defy all historical precedent. Trump can only lose in 2020. We cannot defeat him if he doesn't give it away. That's the sad situational reality.
The good news is that he has done a marvelous job of giving it away since early 2017. But undecideds and swing voters were always going to trend back to the incumbent. I have predicted for years that his approval rating would uptick toward 45% level on 538 as the 2020 election approached. Hispanics in particular have an overwhelming tendency toward benefit of a doubt to the incumbent. Trump's lies and grotesque personality and all the insults are now an accepted part of American life, whether we want to believe it or not. The shock factor has dissipated immensely. That is another aspect I forecast in 2017.
If we continue to disbelieve any poll that shows Trump's approval trending upward, it's going to be a long pathetic year. Keep in mind that his actual popular vote percentage will likely be at least 3% above the 538 number, i.e. 43% approval equating to 46% of the vote, and so forth. I think we can defeat him at 46% popular vote like 2016, but not at 48%. That's when our electoral college deficit gets in the way. It is the reason it is so vital to keep Trump's approval where it is now, and not another 2-3 points higher.
We desperately need him to make a series of atrocious gaffes, like something bigoted or so unkind that even people who had been trending back in his direction are shocked and retreat.
The other possibility is an economic collapse and specifically the stock market. I am heavily involved but more than willing to accept that scenario if the timing is proper. Unfortunately many forecasts have it delayed until post-election.
dustyscamp
(2,224 posts)NewJeffCT
(56,828 posts)if you wanted to stick with an average around the population as a whole?