What is UP with Nate Silver these days?

[AverageJoe90]

Last edited Sat Oct 13, 2012, 01:57 AM - Edit history (1)

If you haven't seen FiveThirtyEight lately:

http://fivethirtyeight.blogs.nytimes.com/

Something is definitely wrong here. Romney's debate bounce ended a few days ago, and Obama is still dropping?!? I've come to notice that Nate's been giving more credence to right-wing pollsters lately as well, so it makes me wonder just how much it's messing with the results. I know he's been heckled by far-right nutjobs in the past, but he's gotta stop giving these obviously biased pollsters any credence because it could start demoralizing people and we don't need that right now. We already saw what happened when the media starting pushing the "Romney won the debate" narrative.

We may need to start flooding his e-mail pointing out what's been happening and why it's screwing up the models, so hopefully, he can get around to fixing the problem.

[hrmjustin]

1. he is using really bad polls.

[Lex]

35. Isn't he using the same polls he always has?

[hrmjustin]

44. he is using these polls that have been shown to be gop leaning ARG in particular.

[Lex]

48. If his model hasn't changed from when Obama was ahead

then it's hard to complain when it shows Obama is now struggling.

[Qutzupalotl]

54. That's only true if the polls keep the same methodology.

If firms like Gravis gradually begin oversampling Republicans, that would skew Nate's results.

[femmocrat]

2. I have been wondering that, too.

After months of having the President miles ahead in the electoral college, all of a sudden the race is this tight? After one so-so debate performance???

I just don't believe it. Maybe the Kochs got to him.

[AverageJoe90]

22. I seriously doubt he's been compromised, though.

It just seems he's been using too many substandard polls. And TBH, not all Republican-leaning pollsters are Razzies, I think.

[Godless in Seattle]

50. Righties have been pumping out polls a mile a minute since the first debate

If Nate is taking those into account, naturally his numbers are going to skew toward Romney. But it is a legitimate question to ask WHY he's taking these polls we've never heard of into account.

[chemp]

3. e doesn't write the polls

He interprets them.
Rmoney is going to sink again. Count on it.

[gkhouston]

4. He's not interpreting them. He's plugging them into his model and interpreting that.

The question is: should he be taking a more jaundiced view of some of his input data?

[AverageJoe90]

18. You put it better than I did, thanks, GK.

[AverageJoe90]

25. Hope so. The undecided voters can't be as swayable by B.S. as we may think.....or can they?

[blueknight]

5. pres blew the debate

that simple! you dont have to be a pollster to see how much damage that did to us. he lokked un-prepared, un-informed, and basically, like it didnt matter to him. when something matters to you, you will fight like hell. pres didnt fight at all. Now he's paying the price, as all of us are

[AverageJoe90]

16. Somewhat true, but Romney's bounce ended not that long ago. nt

[Cha]

58. You don't have to be a genius to see how

the corporatemedia excused romney's lying and bully and declared mitt the winner. And, the left media except some levels heads raged on and on about the President too. Which made the gopropaganda machine SO MUCH EASIER.

Don't worry blueknight there are two more debates and the Pres has learned his lesson. He's had my Back for four years and I want him to have it for Four More.

I have his back, too.

[arely staircase]

6. nate's numbers are as good as they were when they had us winning

no more

no less

let's get to work!

[AverageJoe90]

12. I'm not so sure of that anymore.

I knew this race would tighten. It's given. But even under the circumstances, it doesn't seem quite right, and I'm usually right about stuff like this.

And yeah, I agree on one thing: Keep working.

[arely staircase]

17. im not sure either

but i am writing another 25 dollar check to the obama campaign.

i am a texas teacher and that is not insignifigant - i have given 300 to date.

[AverageJoe90]

24. I hope to start donating myself. Wish me luck. nt =)

[arely staircase]

27. luck upon you

[LeftyMom]

7. Shooting the messenger?

Everybody thinks the guy's a genius when he's telling them what they want to hear.

[DisabledAmerican]

13. what happen to having a different point of view?

Come on now we are a democracy now. If someone is proving his model doesn't work because he is plugging in outlier polls into his model you could be just as bad on shooting the whistle blower. Two right wing polls are already messing up his models.

[LeftyMom]

15. Remember a few weeks ago when the right was all about skewed polls and creating their own reality?

Don't be those guys.

[AverageJoe90]

19. I agree, but that isn't what I'm doing.

As GKHouston so correctly put it, shouldn't Nate be looking at some of the sources?

[DisabledAmerican]

23. Your concern is noted seriously though I'm not buying the concern

If Nate is putting bad info into his model that effects his model. People can look at the info that is not the same. You know this whole we can't look into the polls memo cause of freepers is laughable. Really should try harder then that if that is your only concern.

[LeftyMom]

28. Are half of your 300 posts accusing people of concern trolling?

My point is that Nate Silver is an expert on polling and the people who think they've found a problem with his work are not. This was ridiculous when the right was doing it and it's ridiculous now.

[DisabledAmerican]

32. Do you even know who the people are that found problems with his info?

You accused people of being as bad as freepers just for looking into what polls are being used in his model. Two right-wing polls were noted tonight as part of what he entered into his data. You do not think that is alarming that he is putting well known right-wing polls into his model?

[LeftyMom]

36. He adjusts for biased polls. This is not new.

Take a deep breath and stop freaking out.

[DisabledAmerican]

42. who said I was freaking out I just said this should be entertaining let them look into it

[DisabledAmerican]

60. I'm not the one concerned about a person looking into

what is going on with Nates Model. I look at it this way it's all about improving the system. Any data that helps improve the system is a good thing or are people afraid of looking at info to improve it? Actually that I do find interesting don't you?

[AverageJoe90]

14. What you've said is true, in a way, BUT.......

There's a difference between shooting the messenger, and asking if perhaps Nate's models are being fouled up a bit by bad information.

[LeftyMom]

20. If they are it'll clear up in a few days with new data.

The only poll that really matters is the one in November.

[DisabledAmerican]

33. Bingo which is why do not worry if they look into it.

The only poll that matters is Election results lets go Get out the vote and get Obama a win.

[AverageJoe90]

37. Maybe, but maybe not. We'll see. nt

[Cha]

59. Or maybe Nate's not perfect? Maybe he is figuring in bad polling data? nm

[DemocratSinceBirth]

8. You Don't Have To Be A Master Aggregator To Know The President Isn't Doing As Good As He Was

eom

[DisabledAmerican]

10. Your concern has been noted

[DemocratSinceBirth]

30. What Did I Sat That Is Remarkable?

Nate isn't some genius who God dropped on us lesser beings to lead us to a greater understanding of the ebb and flow of a presidential election.

[DisabledAmerican]

9. This is going to be interesting

Anyone want to pass the popcorn? It's about time people start to wake up.

[amborin]

11. disagree

Nate discusses pollsters in depth. Read some archived articles. He weights pollsters according to house effects.

That said, Nate himself cautions about reading too much into some of them.

[budkin]

21. I think it's because the debate screwed his numbers so badly

That he is starting to second guess himself so he's incorporating all the wingnut polls into his estimates.

[AverageJoe90]

26. I'm beginning to think so, too.(no conspiracies, though)

I don't really blame him, since a lot of the right-wingers were heckling him for so-called 'bias'(when none existed!), but it's starting to affect the quality of his models. If he can't fix the problem then we may have to start looking for secondary sources for a bit.

[ProSense]

29. Here's the problem,

from the current piece:

It might be noted that some of the state polls released on Friday were from firms that have had Republican-leaning results, and they might slightly exaggerate his standing. It is unlikely, for instance, that Mr. Romney would win New Hampshire by four percentage points right now, as implied by an American Research Group poll of the state.

http://fivethirtyeight.blogs.nytimes.com/2012/10/12/oct-12-romney-debate-gains-show-staying-power/

A lot of right-leaning polls dragging down the averages. It's impossible to get an accurate picture of the race from mostly right-leaning polls.

[DemocratSinceBirth]

31. How Come The Only Pollster "We Have" Is PPP And The Pubbies Literally Have At Least A Dozen?

.

[regnaD kciN]

46. Three gue$$e$...

That is a real problem -- through most of this election season, you'd only occasionally get a dicey Eepublican outfit poll on an isolated state, and maybe, once a month, one such organization would dump (appropriate word) five or six skewed state polls at once, but that was about it. In the last week or two, it seems like you get such a dump every day -- or maybe two such firms, covering different sets of states (which, curiously, seem to be ones where a more-reputable firm has just reported "safe" numbers for the Democrats). There's just so much additional noise out there, it's hard to separate the wheat from the chaff -- and it seems as if Silver's solution is to abandon trying to draw such distinctions and just treat everything as being of equal significance.

[andym]

34. It's the inertia of political events that are affecting the polls-- but it can and will change

Yes indeed. Nate is just responding to the numbers.
I'm very hopeful that by this time next week things will be looking a lot better.

See here about political inertia.
http://www.democraticunderground.com/1251137992

[DemocratSinceBirth]

40. He's A Political Weathervane.

Nothing more and nothing less.

[AverageJoe90]

45. I took that all into account before writing this, though. nt

[Coexist]

38. Nate's a numbers guy

he tells the truth as his analysis sees it

[tritsofme]

39. I won't believe him again until his numbers look better for Obama.

[ProudToBeBlueInRhody]

41. It's not like Sam Wang is any more positive in the numbers

http://election.princeton.edu/

And he's was saying Obama was a lock.

[unblock]

43. i think his model doesn't just aggregate a day's polling. i think it extrapolates a trend.

i think, to some extent, if it sees an increase in the polls over some period of time, it assumes that it will, to some extent, continue. so i think the 80+% numbers we saw for a while were to some extent based on an assumption that obama's rise in the polls would continue.

when the polls reversed course after the debate, obama's number came down. after a few days, silver's model identified a trend for rmoney, and obama's number got hammered as it assumed rmoney's rise would continue.


i'm not discounting problems others have noted -- overweighting of suspect polls, in particular -- but i think this trend aspect exaggerates the problem.

[ejbr]

47. I would rather think that

Nate is off than people are actually buying the shit spewing from the other side, but after Bush, it wouldn't surprise me about the stupidity of the American public.

[Samantha]

49. I have been wondering the same thing

Last edited Sat Oct 13, 2012, 09:43 AM - Edit history (1)

I started a thread here on September 29 which was entitled Nate Silver gives Obama a 69.4 percent chance of winning Florida. On that day, he also predicted:

319.3 Electoral College Votes Projected for Obama
218.7 Electoral College Votes Projected for Romney

I checked earlier today and saw

283.1 EC votes for Obama
254.9 EC votes for Romney

66.8% chance of a Romney win in Florida (can you believe that after what he predicted on September 29th?)

57.4 chance of Romney win Colorado

52.5% chance of Romney win in Virginia (it was 76.5% percent chance of an Obama win on September 29)

63.5% chance of Obama win in Ohio (it was 83.9% chance of Obama win on September 29th)

89.5% chance of Obama win in Pennsylvania (it was 97.1% chance of Obama win on September 29th)

92.7% chance of Obama win in Michigan (it was 96.8% chance for Obama win on September 29th)

75.5% chance of Obama win in Wisconsin (it was 89.7% chance for Obama win on September 29th)

I simply cannot believe these dips in these percentages in not quite two weeks. It is just too astounding. Even if one believes Romney won that first debate (I personally do not; I think he defaulted through cheating and lying), that to me does not explain these steep drops.

If you have any thoughts on this, I would love to know what they are. I am, BTW, a Nate Silver fan. But something here does not pass the smell test.

Sam

September 29th thread: http://www.democraticunderground.com/10021438787

[AverageJoe90]

61. Thanks, Sam.

Honestly, I just think it's a case of an overdose of bad polls and poor sources at the moment. But it is a little concerning because we both remember what happened after Obama supposedly 'lost' the debate with RMoney.

However, though, I do hope enough people will point out the irregularities, and that Biden's kicking butt will have a positive effect on the polls.

[cecilfirefox]

51. For the most part, I trust him. You realize the right wingers have folks saying the same thing? nt

[AverageJoe90]

55. Not quite the same thing, Cecil.

Here's the difference: The right-wingers were actively screaming about skewed polls. All I'm saying here is that Nate seems to have gotten a tad careless with what sources he uses to calculate the race's outcome lately, especially since Romney's bounce is over.

[Logical]

52. LOL, once again, we love Nate when it looks good and hate him when....

It doesn't! And do you REALLY think Nate gives a shit what emails say???

[AverageJoe90]

56. Here's the thing, Logical. I don't 'hate' Nate.

TBH, I was disappointed when 2010 rolled around, but as far as I know, there was no problem then. The truth is, the quality of the modelling has indeed slipped a bit since Nate started including more right-wing polls.

And yes, I think if enough people pointed out the truth, Nate would certainly be willing to reconsider......I think he would.

[Logical]

67. Nate know more about right wing polls than the whole DU combined....

He has no bias. He tells it like he sees it! And if emails change his mind he needs to be fired!

[AverageJoe90]

68. Yeah, but it certainly doesn't invalidate my point.

It's not about bias. Believe me, I know. But somewhere along the line, he screwed up somewhere and started letting all these rightie pollsters in the door. I'm sorry, but like I said, it doesn't add up, and you are naive if you think otherwise.

[Logical]

71. LOL, maybe Nate was wrong when he had us winning 80% to 20%. I would love to match your...

qualifications again his.

Post a link to some of your analysis. I would love to read it.

[AverageJoe90]

75. C'mon man.

TBH, I've never been all that good at explaining things so maybe you can do a little searching for yourself.

I don't have a link on hand but someone on DailyKos pointed out that pollster Mason-Dixon has had some issues in the past, mainly with demographics and stuff.

[Logical]

80. Nate NAILED the 2008 election and predicted we would get our ass kicked in 2010 and the DU HATED...

him in 2010.

I will trust him until he proves he made an error. So far he has not.

[AverageJoe90]

81. Very true, BUT......

I don't think the model had any problems with excessive numbers of right-wing pollsters having their shoddy data thrown into the mix 'till just these last couple of weeks. Nate DID make a mistake there, no question about it(though he may be fixing it, though).

[MrSlayer]

53. We should complain because we don't like the polls?

It never occurred that maybe we're losing ground? That the abject fail that was the first debate did serious damage? That people in this country are dumber than a bag of hammers?

Perhaps what's up with Nate is that he's doing his job.

How do we turn it around?

[AverageJoe90]

62. We may have lost ground, but.....

A lot of people I've come across, particularly on DailyKos, have concluded that Romney's bounce ended a couple of days ago. And yet, Nate's odds are still going down. Something is definitely off, and fellow DUer Samantha made the point that it doesn't make sense to have this much of a serious drop in just two weeks; Obama may have been slightly off his game, but not enough to justify what we've seen in the 538 tracker.

Everything just doesn't quite add up. It doesn't take a conspiracy theory to come to that (logical) conclusion.

[muriel_volestrangler]

65. What makes you think Kossacks can know a 'bounce' has ended?

Which are the polls that have shown the bounce is over?

Isn't it worth pointing out Silver still has Obama at a 61.1% chance of winning?

[AverageJoe90]

70. It's difficult for me to explain right now.

This will help explain things, though:

http://www.democraticunderground.com/?com=view_post&forum=1251&pid=139075

Yep. And frankly, if it weren't for those crappy right-wing polls Obama's chance for winning would likely be about 70% right now....and frankly, that's a little pessimistic, TBH.

[cali]

63. Sorry, I think Nate Silver knows what he's doing.

[AverageJoe90]

64. I'm not really saying he doesn't, BUT......

Cali, if you've regularly watched the site, as I have, you will have noticed that he's given a lot more weight to right-wing pollsters, many of whom are pretty obviously biased, and it's starting to mess up his models. Even a smart guy like Nate can make mistakes, and hopefully he starts fixing them soon.

[byeya]

69. I agree with what you are saying and my guess is that Nate is in the process

of damping down the RW noise in a methodical manner right now.

[AverageJoe90]

76. I would hope so.

I just wasn't sure I could trust his modeling 100% simply because of the anomalies I, and others noted. Hopefully, Nate really is adjusting where needed.

[Turborama]

66. Recommended for the interesting debate this OP has generated n/t

[AverageJoe90]

73. Thanks.

I do feel this was an important thing to point out. Frankly, I don't any see conspiracies at all: Nate's a decent fellow, ya know!
All I'm saying is is that some of his sources seem to be weighing down on his models a little too much and that I think he should reconsider some things.

[Turborama]

85. Looks like your theory has been validated here...

http://www.democraticunderground.com/10021542700

[former9thward]

72. You clearly know how to interpret polls better than Nate.

He clearly needs your help since you know all there is to know about polling and what polls to use. You should apply to be his partner. I'm sure he will take you on.

[AverageJoe90]

74. Again, the problem is some of his sources, such as Mason-Dixon, etc. n/t

[former9thward]

79. No that is your problem not his.

He uses polls which skew to the left and skew to the right. He balances them based on their history and current turnout projections. You just want him to use left leaning polls. With polls I would rather see a best guess at a correct number rather than what I may want to see.

[AverageJoe90]

82. You're not listening, man.

Again, my problem is, he's been using way too many right-leaning polls, and many of them with suspect data at that. That was my issue.

[OhZone]

77. gigo nt

[AverageJoe90]

83. Explain, please? nt

[OhZone]

84. G.I.G.O -

Garbage in - Garbage Out.

If he's being fed bad polls, then his statistics will be damaged.

Oh well.

[CoffeeCat]

78. I heard (from all places, Fox News) last night that the latest Virginia poll...

...was conducted by a well-known Republican pollster.

I haven't heard this from the MSM.

I think the major polls are reliable for the post part. However, I do think that Rasmussen is often used to numerically cement a meme the right wants validated. They're right most of the time, otherwise they would be unreliable. However, you can see when there are shennanigans happening.

We've be naive to believe that polls aren't used to manipulate.

When it comes to all of these unknown polls that suddenly crawl out of the woodwork--and have disproportionate gains for Romney--yeah, I'm suspicious.

But again--I do have confidence in well-known, established polls that have been around for a while.

I'm feeling better about Virginia--since learning on Fox (!) that a Republican pollster conducted the latest poll that shows Romney ahead by 2.