## I Don't Get It

I refuse to follow the ins and outs of polls and the horserace aspects of elections.  But I couldn't miss all the blog activity that somehow Nate Silver is purposefully corrupting his election predictions for some partisan reason.

A physics professor once used to tell us that if we don't even know the sign of the answer, then we should assume we have no understanding of what is going on.  Well, I don't even know the sign of the answer here.  Would a partisan inflate Obama's predicted chances of winning, thus giving him some sort of momentum?  Are there voters who just want to be on the winning side and vote on election day for whomever they think is going to win?  Or would a partisan make his man look worse in order to panic the base and make sure they get out and vote?

1. #### Mark:

Here is the problem with Silvers "analysis": it is impossible. He is claiming statistical odds of Obama winning when there is no significant statistical method to make those assumptions. And, his defenders, like Paul Krugman are trying to pretend that his analysis is some form of science. I think I read Silvers blog one time and he claimed an 83% chance of Obama winning. Maybe I misread this, but there is no way any analysis can make these claims. Forecasting election results has long been an interest of mine, and I am actually pretty good at it. But, for as much science as you can pack into it, it is as much an art.

The biggest issue is that the data really does not exist for such robust predictions. You can analyze previous election results but that only gives you 4-5 data points that can be reliably used (I use 4 in my analysis). There are trends and outliers. There are both little trends and big swings in voting patterns. THere are favorite son impacts. There are shifts in voter identification and turnout.

So, Silvers could very well be correct in claiming an inevitable Obama victory. I predict that Romney will win with 50.8% of the head to head vote and 285 electoral votes. But, with that prediction there are 6-7 states that are incredibly close that in the end could swing either way. What I do know is that making any claims to the CERTAINTY of my prediction are false, and anyone who makes such predictions of certainty are fools or liars.

2. #### markm:

"Are there voters who just want to be on the winning side and vote on election day for whomever they think is going to win?"
Yes, yes, depressingly yes.

3. #### Mike Harris:

Mark,

His estimate of the vote share have implied error bars / confidence intervals around them, codifying the uncertainty you discuss. The overall win% number incorporates those error bars.

I understand that in the end we are all guessing, but to be this seems like a reasonable set of steps.

538's model may be wrong, I hope it is. But all the criticism of Nate Silver bothers me. My opinion is that Silver is a "man of science." Amng mainstream journalists, he is easily in the 99th percentile of math / probability literacy. Focusing criticism on him makes me sad.

4. #### DrTorch:

It's a matter of knowing your audience. If Silver is loud and right, he'll get more attention and opportunity now and later. If he's wrong, he may even get one more shot in 2016.
If he's quiet, it won't matter much whether he's right or not. If he's a quiet, cautious scientist, he'll get ignored. The media, and enough of the audience, wants boldness, flair, form over function, style over substance. So being cautious offers no benefit. Bet big to win big.

5. #### Todd Ramsey:

Every week for six months Karl Rove has been writing a WSJ op-ed showing that Romney will win the election because of thus-and-such polls or projected turnouts. It seems unlikely that he has not seen a single piece of evidence over that time indicating that Obama will win. My conclusion is that Rove believes that his op-ed predictions of Romney victory will influence the outcome in Romney's favor.

6. #### sabre_springs_mark:

But the difference is that the polls are in Romney's favor and have been all month, so Rove is essentially correct. What is your issue? If you do a serious analysis Romney is going to get from 290 - 307 electoral votes. I could see early last week Romney would win it.

Now the Sandy storm does put a wrench in things a bit, but I don't think Obama is looking presidential enough to make a difference. But we will know for sure when Gallup releases its first new poll since Sandy - they said they would start again today - I don't know if that means they will release a poll today, or if they are going to start polling again today.

7. #### sabre_springs_mark:

It isn't just that guy. Yesterday the Quinnipiac Poll released numbers showing Obama up +5 in Ohio - but they are assuming +8 Dems over GOP - basically what occurred in 2008. They can only be doing this to try to throw the figures to give Obama momentum. Fortunately a "serious" poll was seriously debunked when folks dug though the numbers. Even NPR only used +4 Dem, and the way the election is going it looks like the Dem advantage will be -1 (per Gallup)

So forget this guy's blog is off, why is Quinnipiac risking their entire reputation on what has got to be known as a fake poll?

8. #### sabre_springs_mark:

Even serious polls always poll wrong. PPP for instance. I mentioned Quinnipiac above, a total piece of junk released yesterday estimating Dems at +8 when everyone else has at least gone to +4 and the two biggies are using even or -1 Dem advantage. You would think they have a reputation to uphold (PPP is an admitted Democrat propaganda poll - not kidding about this. even RCP lists it as such but uses the figures anyway)

9. #### sabre_springs_mark:

And before you write me off as a raving Romney loonie - I voted for Gary Johnson on Warren's advice actually. I am just a severe poll hound. Though even I am getting tired of "Bronco Bama and Mitt Romney", though it doesn't drive me to tears.

http://abcnews.go.com/Politics/OTUS/viral-video-girl-cries-bronco-bamma-mitt-romney/story?id=17610152#.UJKBu2_A9Cg

10. #### sabre_springs_mark:

This close to the election you can actually get some kind of scope of who is going to win - though there have been last minute surprises, like when Bush was 4 points ahead and the media in the last weekend hounded us over a drunk driving conviction of 20 years prior. (Was that really news? I guess it was more important than ambassadors dying - A for the quaint ots when we were still so innocent that a DUI from 20 some years ago was still a big issue.) It was enough to drop the lead from 4% to 0.

Looking at the current polls and adjusting for biases, assigning the undecideds (75% against the incumbent it looks like Romney has an easy win. I am also guessing GOP will get 51 in the Senate - Dick Morris says 53 - I think Dick might be smoking something. But then is Sandy the last minute surprise? We don't know yet.

11. #### sabre_springs_mark:

Something else I just remembered about Carl Rove. Just two months ago he was predicting an Obama victory. So just dismissing the guy because you don't like him and then complaining that he says his guy is going to win, when the polls are strongly in this guys favor (at least the ones not polling Dem +8 and there are a few) is being strange on your part.

12. #### obloodyhell:

}}} Amng mainstream journalists, he is easily in the 99th percentile of math / probability literacy. Focusing criticism on him makes me sad.

I *AM* in the 99th percentile of math/probability literacy. Not just among journalists. If he's giving Obama a 75% chance of re-election, he's high on drugs, not math. I'm not certain Obama will lose, but I can pretty much guarantee a pro-Romney landslide is more likely than a pro-Obama one.

13. #### Will:

Probability is not the same thing as certainty.

14. #### sabre_springs_mark:

Seems like as good a method as any. Is it better to vote for someone because of your extreme loyalty to a party, or because the Union says you should or ...? Seems like if a lot of folk think X is going to win - X has greater chance of being a better president.

Hate to say this, but voting for "the winner" as the better candidate has worked in the last oooh - 4 elections or so. And yes despite his flaws, Obama was probably a better choice than McCain. And say what you want about W, you really Kerry would be better? How 'bout that Bob Dole over Clinton ...

This is not to be confused with the phenomena of increased perceived votes over time. Reagan in his first election won, but not by that much, by the time his second term was about to begin, it was pretty hard to find a person who "didn't vote" for Reagan in 1980.

15. #### sabre_springs_mark:

Something else I find interesting. We all hear about within the margin of error. Well the margin of error is considered 2 standard deviations from the norm- or approximately a 95% confidence level. 75% is well within the margin of error - will that be the claim when the prediction is false? "It was within the margin of error. "

16. #### sabre_springs_mark:

Admittedly taken from a partisan source - but the table accurately reflects the polls. http://www.nationalreview.com/corner/332170/turnout-game-josh-jordan#

Double click on the picture to see the full poll results vs Dem / GOP turnout assumption - the results are crazily directly related to the assumption. Just guess how much +- you think Dems will show up and you will ascertain the results.

17. #### Dan Sherman:

Doesn't matter. The end result will be the eating of some tasty crow, by Mr Silver. Of course, he'll just say it ended up being the 25% that won out (of his "statistical" model.)

I'm not sure what I'm looking forward to most on Nov 6th. Romney's win because the future of our country just MIGHT get marginally better or Romney's win just to watch the libs excrete themselves.

18. #### Dan Sherman:

One hole in your theory is that in order for a pundit to try to shout their way to a self fulfilling lead, they have to be covered by mainstream news media in order to get that information to the left in this country. But I can assure you, not one "on the street" liberal could site to you what Rove has been saying for the past 6 months. That kind of blustering your way to a certain outcome only works on the left because they have a MASSIVE media machine that will echo their every word until it becomes gospel... despite the actual truth.

19. #### Dan Sherman:

BTW... my prediction.. Romney 330+. Based on my own mathematics. Obama x 4 = 0 (Remainder is < 270.)

20. #### Stan:

My poli-sci professor told me Americans tend to agree with the majority, when told what the majority is, even if it isn't actually the majority opinion. I'm sure there is some science to back that up, but I could care less.

21. #### Nehemiah:

There may have been a time when polling had some credibility. However, in this day and age caller ID allows us to screen out those unwanted callers (pollsters). Who are these people who actually participate in these polls? Mostly get a lifers I suspect! Your civic duty is to inform yourself and then vote, not participate in polls pre-election or upon exit from the poll.

And of course there are political reasons for fudging poll numbers up. It sets up a narrative. Say Obama was way ahead in the polls going into the election. How could he lose? There must have been voter fraud or suppression. The teams of lawyers are already geared up and ready to go. Anecdotal evidence will be ready. Eye witnesses have been rehearsed. Yep, high poll numbers for Obama in Ohio give credence to a "they stole the election" mantra. Good thing we will have those UN observers onsite to monitor the results. They will certainly be unbiased in their evaluations.

The 2000 election will seem like a cake walk compared to 2012.

22. #### Steve Walser:

A true "man of science" would not state the probability of an event as unpredictable as human decision making to the 1/10th of a percent exactitude as Mr Silver does on hundreds of races across the US