Anyone who follows data seriously is continually amused and angered by its use. From Booman citing of a Loras Iowa poll that virtually everyone knew was bad, to this morning’s MD poll of Florida.
So some history. Mason-Dixon did polling on Florida for the St Pete Times. Their last poll had Romney up 6.
That poll had an effect on people. The St Pete Times is well regarded, and their poll suggested Florida was gone. I well remember the Obama campaign reacting saying they KNEW the number was wrong.
Except MD was way off. So Adam Smith and the St Pete Times staff reviewed the results:
AND FIRED MASON-DIXON.
So MD is back at it this morning. But of course the reporting on it makes no mention of MD’s miss in ’12.
No accountability. Not even any attempt to ask the question.
Which is the way it is in politics generally.
MD has long been so bad in polling that whenever I look at polling numbers, I toss it out completely. Not even worth including it in the aggregate trends.
I don’t know if MD is high quality on the nuts and bolts in collecting its samples and running the data through standard statistical methodology like Loras but still gets it all wrong or if it’s sloppy at every point in the process. Bottom line is that both outfits aren’t getting any more accurate. Unlike Q that appears to be improving with each election cycle.
IMO this article summarizes what’s wrong with political polls. A worthy quote:
“The problem arises when thousands of public figures and marketing departments launch their own surveys to gain competitive advantages. The result is an over-saturated public-opinion marketplace in which citizens and customers alike begin to feel a mixture of fatigue and irritation that dilutes the accuracy of the polls, leading their trustworthiness to get called into question.”
http://theweek.com/articles/559473/how-did-political-polls-become-inaccurate-blame-obsession-data
In this election I have received more calls than ever before. I have the feeling my opinions are being shaped by the questions the interviewer asks. Also, there are too many questions and the call lasts typically 15-20 minutes. How many people are going to participate in such a lengthy interview? I felt exhausted after the experience. Additionally, there are too many options to respond to, but a “neutral” was not one of them.
Funny thing. People think polling is better now than it was. There is more of it. But there is far less rigor in it than there was 40 years ago
Gallup went door to door. They had a refusal rate of less than 10 percent. Refusal rates now exceed 80 percent
The is so much junk. I am suspious of Internet polling though ypugov seems to have figured it out
The best pollster over the last few cycles is pew, but fox cnn abc and cbs are good. Ppp has been good. I am suspicious of Marist Suffolk and Monmouth. They have had a good cycle or two but their history is short
In general the article you cite.
I’m sceptical about yougov. In the European parliament elections in Sweden 2014. The Feminist initiative secured an early position as likely to gain seats by early showing in Yougov polls. We in the Pirate party first sighed because everyone should know that Yougov is a glorified internet vote, and then slapped ourselves in the forehead and registered a bunch of accounts.
Come next poll, Pirates had a decent looking 6% (enough to defend at least one seat), in contrast to around 2% in other polls. Unfortunately the Feminist did not settle with decent but had an unrealistic 18%. Which lead Yougov to change the results after publishing, blaming computer error. An error which changed the result in just Sweden, and only for two parties, one of which I know for certain registered accounts to vote up itself, and one where I have a strong suspicion of the same.
So you are right to be suspicious, but wrong to give Yougov a break. If they changed the results in 2014, what is to stop them from changing results other times if it looks to different from other polls? And then they are worse than useless.
I, too, am skeptical of Internet-based polls, but just because you register an account doesn’t mean you can inflate your chosen party’s share. That’s why YouGov asks a whole bunch of other questions — so that they can take a stab at choosing a pool of voters that matches up well with the demographics of the LV population.
Of the online pollsters, YouGov is one of the few ones who can be classified as a serious pollster. They miss a fair bit, but they also get a fair bit right.
Remember, too, that polling European elections is not the same as American elections. As I understand it, when you have an election with n > 2 parties, things get a bit dicey. Combine it with the inherent difficulty of overcoming the sample bias in Internet polls, and polling (say) Britain can be a fair bit tougher than the US.
On the internet, you may be an old conservative if you wish to. And if I know that they at one time fudged the numbers and blamed computing error, I get a sneaky suspicion that the reason they “get it right” might be that they fudge the numbers if it looks wrong.
But I will give you that it is much easier to rush a poll in Sweden then Britain and easier in Britain then the US from shear numbers. And two party versus multi-party sounds reasonable.
I don’t need you to give me anything, bud.
I am sorry, but I must be missing your point.
Eh, I assume you’re referencing the FL poll from yesterday that had her up by two. That’s not way out of line if you think it’s a four- or five-point race, nor would it be if it had her up eight-ish. Plus, whiffing four years ago doesn’t mean whiffing this year.
They’re not great though. Not schizophrenic like the Q-poll (whose Florida polls swing absurdly despite the narrow range of outcomes Florida is predisposed to), but nobody’s ever going to confuse them with Marist or Pew (or even PPP/SurveyUSA-types).
Pollsters miss. It happens.
But most people who write about polling regard it as a way to prove something they already believed. Coker of me is a Republican. He spins his polls. When he does that he should be asked about 2012. He said Florida was over in 2012 and Romney had won it
Have you seen the surprise results being reported from mail-ins to the FLA primary? Lot of unpolled voters showing up.
This has a good summary.
http://www.tampabay.com/news/politics/stateroundup/ballot-is-in-the-mail-for-near-record-22-million-
florida-primary-voters/2286911
Five weeks before the August 30 deadline, the mail-ins already almost equaled the number of people who voted in the 2012 primary. About 45% of the ballots were Republican, 37% were Democratic; the remaining 18% were not party affiliated. Although the primary is closed, people can still vote for some races.
He should be if he’s going to spin, but like all pollsters, he’s just trying to generate publicity, I suspect.
I’d add that, in re: accountability, the St Pete Times did fire them after their 2012 miss.
And the arrogance isn’t nearly as hilarious as (I think) Suffolk was back in ’12. Didn’t they quit polling Florida in September or something and declare Romney had it wrapped up?
Emerson College
As Billmon said,