Politics, polls and intellectual dishonesty - SHOCKED!

[RoyalShock]

I came across this graphic being shared on facebook. When I saw the total respondents number my first thought was that based on my recollection, that's pretty much a typical number for national polls I've seen in the past.



So I decided to do a little bit of research and found that for a population of 150,000,000 (estimate of actual voters):

- A sample size of 385 will give you a confidence interval of +/-5% 19 times out of 20 (95%)


I'm going to presume this particular poll wasn't just to identify "Total Republican primary voters", but probably to survey Repubs, Dems and independents, in which case you can't really expect more than the 1289 total respondents. The weighted number of 355 isn't out of line with most polls where there is a +/- 5% margin of error.

I know the focus of the graphic is how the weighted number was arrived at. But if the Trump campaign is going to dismiss one poll for weighting, they shouldn't tout any poll at all. So who is really doing the misleading here?

Source: http://www.raosoft.com/samplesize.html

[SB Shock]

From looking at the image u posted - I think he is complaining about how they manipulated the poll - actual respondents that were republican was 44% (575 out of 1289), but the pollster decided to weight them down to 25-27% (assuming the total was 1289 or 1331). Of course - the model the pollsters uses to weight the respondents is always key to how accurate the polls will be. In the past when Conservatives have surprised it was because their models were wrong.

[MoValley John]

I can't stand Trump. That said, plus or minus five percent is a huge margin. Huge! You can't trust the polls, any of them. Gallup was so embarrassed about their inability to accurately poll primary elections, they no longer do it.

[RoyalShock]

From looking at the image u posted - I think he is complaining about how they manipulated the poll - actual respondents that were republican was 44% (575 out of 1289), but the pollster decided to weight them down to 25-27% (assuming the total was 1289 or 1331). Of course - the model the pollsters uses to weight the respondents is always key to how accurate the polls will be. In the past when Conservatives have surprised it was because their models were wrong.

Yes, and I said as much about the weighting being the focus of the graphic, even though the bolded title focuses on the 355 number not being "national" enough. All polls are weighted. If they knew how the weighting was done, they could point out exactly why they believe it's misleading. Instead they use similar tactics to do their own misleading.

[SB Shock]

I just read it differently that the focus is on the weighting - which is suspect since 35% in 2012 were republicans and now it being weighted to 25-27%. But of course I would want to see the actual full poll report because some of the stuff doesn't add up in the trump graphic.

[wu_shizzle]

https://www.conservativereview.com/commentary/2015/10/the-biggest-scam-of-the-2016-election

Even Gallup doesn't have a lot of faith in their own polls.