2016 Postmortem
Related: About this forumDamn it, I guess I have to do a "Warn Off" for Overtime Politics (Ignore them, entirely)
So, Overtime Politics is showing up all over DU these days because they're producing numbers way different than anyone else and that can be fun to see... especially when it benefits your candidate. However, Overtime Politics is not a real polling outfit. Not even close.
They really first showed up on the DU radar when they posted a poll about Bernie being within twenty of Clinton in South Carolina. The no longer with us poster who started that thread vehemently defended Overtime because.. well, I'm not really sure, to be honest.
But Overtime had two attractions going for it:
- They're producing contrarian results
- They assign an MOE to their polls, which make them look valid.
So what is Overtime?
First of all, they're brand new. I mean really new. Too new to even have an operation in place to produce valid polls. The site was registered on October 20 of this year, out just over two months ago.
However, they're planning on hiring a second phone operator to be able to release 3 or 4 polls per week per race. That, well actually that is scientifically impossible to do. The pace at which they release polls should be a huge warning flag.
But maybe they've figured all this out. Maybe they've managed to create a streamlined process that will be the envy of actual pollsters. Well, they have streamlined something...their methodology:
"We call both landlines and cell phones at about a 3:1 ratio, however, telephone polling is a very time consuming process, the overwhelming majority of people do not even answer their phones when we call. In order to speed up the process, and get what we feel is the most important metric (Who are you going to vote for?), we only ask our audience if they are planning on voting in their states primary, and if they are, who they are going to vote for."
There is a lot wrong in that paragraph, enough to make the statical analyst in me to just weep softly into a beer. But the highlighted area is of greatest concern. Every poll is only two questions: are you voting in the primary? Who are you voting for?
The reason no one can see demographics in their polls is because they don't collect them. They don't weigh responses, they don't establish population parameters, they don't establish location densities...they don't do any of the very, very basic things real pollsters do in order to establish validity of results.
So what do they do? They tally up the total "votes" for each candidates, divide by the overall total and call it a representative poll for a total population. Since it is missing literally every single piece of work that goes into a poll to make it valid, I can state emphatically that the only thing it is representative of is that the owner of Overtime Politics doesn't know how to poll.
The owner's comment about the lack of methodology and controls doesn't help:
"The site has been getting lots of emails saying our methodology is flawed, but no poll is going to be perfect. There are certainly polls out there with larger sample sizes, smaller margins of error, and weighted demographic averages, and I can assure you that CNN or the Wall Street Journal have more resources than Overtime Politics to fund these polls. We feel as though our strength will be in polling states which do not get much polling attention until right before they vote. Although the polls may not be as thorough as some of the others, we will at least be trying to get some data for you."
And as for the MOE people like to bang on about showing it is scientific? You can calculate one right online. So if he calls, say 400 people and the state has a population of X. Guess what, he can establish a margin of error. It is too bad he literally never established his population parameters so that the MOE actually references something.
And if you still think I have an ax to grind, even though I have always tried to be nonpartisan about analysis, well, please consider that neither RCP nor Pollster have bothered to report any of their findings.
They're not real pollsters, not even remotely close.
You can find the Overtime Politics excerpts at http://overtimepolitics.com/
Be warned, going there may result in a sudden case of the stupids.
JonLeibowitz
(6,282 posts)Fawke Em
(11,366 posts)Just because some polling outfit the OP likes doesn't count Vanderbilt University in their lists doesn't mean it was a bad poll.
Besides, the OP likes some betting aggregate site for judging people.
Godhumor
(6,437 posts)And I hope I have proven that by this point on this board.
I don't like the outfit in my OP because they're a guy calling people, making check marks and calling it a legitimate statewide poll.
Every single poll on that site, regardless of results, should be completely ignored. Entirely. Be like RCP and Pollster, don't give their results even a passing glance.
Response to Godhumor (Reply #7)
Name removed Message auto-removed
Godhumor
(6,437 posts)Not an aggregator of polling. So what it is saying is that Clinton currentlyhas over a 90% chance of being the nominee based on movement within predictive markets. If Bernie wins Iowa, that number will go down some. If he loses Iowa that number will probably stay about the same, as the markets are already expecting Clinton to win Iowa.
KingFlorez
(12,689 posts)Polls generally do not come without cross tabs of some sort.
DemocratSinceBirth
(99,711 posts)EOM
Cha
(297,574 posts)murielm99
(30,755 posts)djean111
(14,255 posts)Tierra_y_Libertad
(50,414 posts)wildeyed
(11,243 posts)Using Nietzsche to counter a poll interpretation you don't like is up there in the logic spectrum with using Bible quotes to defend creationist beliefs. And in any case, he was not arguing against practical facts, just pointing out that confirmation bias exists.
LWolf
(46,179 posts)key-master of pollsters on DU, and, as DU's polling overlord, it's up to you to "warn off" or "accept" whatever poll maker appears.
And, of course, I always take whatever you have to say about polls and poll makers as straight from the poll god's mouth. Because you yourself have no dog in this race, and no personal bias involving primary candidates at all.
Just in case you needed it:
and, edited to add, for the disingenous: I've never seen a poll by the organization you are patronizing us with, and probably never will.
Godhumor
(6,437 posts)BlueCheese
(2,522 posts)Supported by original research, evidence, and thought process.
I have seen polls from that outfit here-- one was even favorable to Clinton.
I guess I have to ask-- why did you respond to the OP in this rather aggressive and snarky way?
LWolf
(46,179 posts)that particular Clinton-supporting poster regularly takes the role of poll-expert on and lectures, and patronizes, those of us who haven't accepted the inevitability of HRC.
My response has nothing whatsoever to do with the pollster; I've never read any of those polls, and am unlikely to do so in the future. It's all about yet another patronizing post..."warning DU off," from that specific DUer.
Aggressive? Yep. Snarky? Maybe. If that's what it takes to make the point, since if I were blunt I'd be hidden.
BlueCheese
(2,522 posts)I find them usually informative.
LWolf
(46,179 posts)through different lenses, from different points of view. I'll stand by mine.
BlueCheese
(2,522 posts)Godhumor
(6,437 posts)And I'll do it without snark or anything else.
I post about polls usually only to address methodology anymore, and even there I almost always do so only to address a misuse or misunderstanding about how something works (Sample sizes, MOE, etc). I rarely address a poll itself directly because I know it can come off as partisan or sour grapes. If you search my posting history you'll see that almost all of my posts about specific poll issues happen only when the result is good for Hillary--it is my way to try and avoid the partisan label while doing so.
This is the first time this cycle that I have addressed an operation directly, and I am doing so because there are fundamental issues with their whole process.
And I don't mean to come off as condescending when addressing this stuff; I act like I know what I'm talking about because it is something I take a lot of enjoyment in by discussing these issues. To be honest, I didn't even think of the OP title as being a "look at me!" thing and seeing what you've laid out I can totally understand how it can be interpreted that way. I actually wrote this in response to a thread from a pro-Hillary poster that blew up with a lot of people asking who are these guys? Can we trust the results?
Look, I tangle it up on GDP, I engage in discussions and I will use snark liberally to make a point. But I try, I really do, to balance that our when talking about analysis. I get that someone acting like an expert can be eye rolling, so I understand if my spiel isn't for you. But know, truthfully, I don't do this to be condescending--I just like it and feel that my background gives me a good understanding of issues surrounding polling. And, honestly, I enjoy writing something that people may find helpful.
Lastly, and this is going to come off as ridiculously smarmy, but I actually respect you a great deal. I used to be a social studies teacher myself, and have found your posts on education to be very informative these past years. One of my only regrets in getting involved in GDP is being on different sides, for lack of a better word, with people I've followed for years. I know of one person I respect quite a bit who has me on ignore, and I'm sure there are others. That isn't fun for me.
So I do regret I've earned a negative reputation with you, really I do, but this is me while I'm active in GDP. When I switch back to GD once the dust is settled, what and how I post will be much different.
LWolf
(46,179 posts)Thank you.
Godhumor
(6,437 posts)I think it needed to be said and talked about.
1StrongBlackMan
(31,849 posts)Based on an extensive knowledge of the statistical analysis.
MohRokTah
(15,429 posts)I immediately dismiss that pollster out of hand.
Godhumor
(6,437 posts)Yea, I wouldn't trust this poll no matter what it showed.
-------
About as much attention as they deserve, in my opinion.
Gothmog
(145,496 posts)randys1
(16,286 posts)is right, saves me a lot of time, what Nate says, period.
fleabiscuit
(4,542 posts)NastyRiffraff
(12,448 posts)Thanks for the in-depth analysis. It helps me (definitely a non-pollster!) understand polling. Keep up the good work!