It takes a pretty smart mind for math to do what Nate Silver does with his statistical analysis, but it’s not like there aren’t a lot of people in the world who can easily understand Silver’s methodology and replicate it. He’s a good writer, too, but not to the point that people read him for his prose. If it’s true that the New York Times tried to keep him and was preparing to give him substantial staff and his own “mini-department,” then they should go out and recruit a statistician who can write and who has an interest in politics. I’m sure someone can be found, particularly since Silver has already laid the groundwork.
His analysis may have irked AdNags and several other bigfoot reporters at the Grey Lady, but it was the most accurate and informative political reporting that the Times produced, which is why it drove so much traffic to the site. There is no good reason to abandon that type of analysis now that we know that it exists.
For one thing, it provided countless people with reassurance when scoundrels like Dick Morris tried to psych us out with their “skewed polls” analysis. There was not one second during all of 2012 when I didn’t know with virtual certainty that Obama would be reelected, and my faith in Silver was the reason why. I probably owe him a couple years of my life that I didn’t lose to stress. The dogged truth, as expressed by Silver, was a powerful antidote to our horrible world of political punditry.
We can’t let something so valuable just vanish. If the Times doesn’t figure this out, someone else will, and they’ll get all the traffic and attention.
Nate Cohn would be a good replacement, but he blathers on too long. Not really web friendly.
I do stats and I do politics and I write, but I’m not sure the NYT would be interested in paying me my replacement cost to leave corporate.
Sam Wang
http://election.princeton.edu/
I’m not sure which of the two I prefer but they’re both fabulous – they don’t always agree on every detail.
Yes, but he already has a cushy and prestigious job.
So did Paul Krugman.
True.
Silver didn’t do anything too special (I got every state right in both elections, which is technically better than Silver, and my vote percentages were closer, and I’m not exactly a statistics master) but reveal that almost every talking head is a fraud, and that said frauds do not want to let the gravy train leave the station; so they kicked Silver out of the caboose and tied him to the tracks.
So yes he is very replaceable. Will they? Iunno. The Talking Heads don’t seem to care for the Gay Wizard(s) interfering with their moneh.
It doesn’t take a genius to predict elections on a widescale level. I got every single state right in 2012, called Willard’s national percentage vote (47% was just too ironic), and predicted Democrats would pick up House and Senate seats.
That said, I do feel like there’s intrinsic value in crossing polls with demographic analysis and past voter trends. It won’t be perfect, though – for example, Silver got 2 Senate races wrong (ND, MT) that he predicted the Republicans winning easily. But there’s value to it.
And for everyone who’s bemoaning his loss…from what I can tell, he’s going to do political commentary (and probably poll modeling) during election years for ABC. If anything, that’s a good thing – his strength is in national elections, not on punditry or even state elections. He gets to do what he really likes (sports) while delving into politics every now and then.
Exactly, it doesn’t take a genius. And it takes a varied higher level of skill to produce a mathematical model to do it, utilizing trends as you stated; this has value.
But it’s very replaceable. But seeing why he seemed to leave kind of leads me to a weird place of something I cannot predict: will they hire some replaceable person who can do (mostly) what Silver did, but also throw a little something to the horse-racers? Or will they just drop it altogether. I feel like it got them too many (much needed) clicks to just abandon it, but the Old Guard doesn’t want to let go of their cocktail parties.
Silver didn’t do anything too special
Josh Marshall had a piece yesterday explaining what Silver did that was so special:
http://talkingpointsmemo.com/archives/2013/07/instrument_controls.php
I just think people are focused on the wrong part. The fact that Silver’s numbers were so good at the very end is not that big a thing. Others came up with pretty much the same stuff. But as Silver would say himself, as his models converge on election day they give greater and greater weight to the actual polls and less and less to economic data, historical data and whatever else he figures into his system. So the fact his model pretty much called it on election day isn’t that big a thing to me; the fact that he pretty much called it six months or 9 months before, based on a system factoring in lots beside polls, is a much bigger one.
Still not seeing it. Didn’t Wang and others have the same thing months in advance? I know I did. I had the same states I picked nine months before the election that I had in election day. However, in between those months I went back and forth on North Carolina and Florida, eventually ending where I began: Obama winning Florida and losing NC.
Now popularizing it and making fools of the Talking Heads is definitely valuable. And his gadgets and gizmos were fun to play with. But on a sheer “who will win” scale months in advance just isn’t that special.
Wang did in particular. Silver and Wang had this in common: they both defined their statistical systems well in advance and stuck to them throughout the election cycle. No mid-cycle “tweaking” to improve results, in other words. They also both had a prediction of the outcome long in advance.
Silver’s model was more sophisticated in several ways that Wang disagreed with. First, Silver incorporated non-polling data and a weighed version of national polls. Wang argues that the non-polling stuff isn’t necessary at best, and at worst it misleads. The non-polling data are derived from past elections and are more likely symptoms than causes. Wang also ignored the national polls as they are meaningless for predicting the electoral college. On these two points I tend to side with Wang.
Another difference was the weighting of polls. Silver had a fairly complex weighting system and included all polls. The weights were determined computationally based on past results of those companies. Wang included the most recent result of most polls but tossed out the known bad ones such as Rasmussen. On this point I tend to side with Silver – his model doesn’t require much in the way of “judgment”. Of course, if the rich wingnuts are able to fund a bunch more polls with consistent reichward skews then Silver’s model could get gamed, so this isn’t conclusive.
One other thing Silver built in was a model of past trends vis-a-vis the key dates in the election cycle -the conventions, debates, and election day. This was actually a very cool idea. I remember that in mid-August 2008 he published an early version of this model showing that if the polling numbers followed past trends you’d see Obama’s get a convention bounce, then McCain, then a gradual return to the mean over the next 4 weeks. This was almost exactly what happened – although in real time most people attributed those polling changes to Sarah Palin and the Wall Street crash. So for 2012 you could see the prediction “if the election were held today” and the prediction on election day itself, which took into account trends. The trend model was, if anything, too cautious. This was a definite relative strength of Silver’s model over Wangs.
But the last big difference probably highlight’s Silver’s model’s biggest weakness. Silver’s model knew, for example, a week before the election that unless the poll results were all wildly off Obama’s chances of winning were in the high 90s – over 99% per Wang’s model at the time. But Silver built in a compensating factor for the poll results all being wrong. Based on analysis of polling and actuals of presidential elections since 1932 Silver came up with a bell curve for likelihood of error in the aggregate poll results and that reduced his confidence level in the results. In 2008 this meant he had an Obama victory at 96% chance. In 2012 this mean he had the chance of an Obama victory in the 60s and 70s until very near the end – and even at that point it was lower 90s.
On that last point Silver outsmarted himself, and probably did so in order to give himself some protection in case his model failed in some way. The reason some of those past elections didn’t match the polls was due to extremely infrequent polls and stone age polling methodologies. With the degree of polling that goes on today the chance that McCain could have won in 2008, given the poll results up through election day, was infinitesmal. The 4% Silver gave for McCain seems small but in fact it says he had a 1-in-25 chance – no way. Similarly, Romney’s chances on election day were better than McCain’s but still well below 1%, as Wang’s model showed.
Well, I think the best measure of this is to look at BP. How has Baseball Prospectus done since he left? I never read it because it was paywalled, but it is as excellent as ever.
In some ways I find the last few years very very weird since I am a big baseball fan and I was aware of Silver since the early BP days.
So what is Keith Olbermann building over at ESPN?
SkyNet.
Has disaster written all over it.
It’s gonna be on ESPN 2. So he’ll be up against The Daily Show and Sportscenter.
Good luck trying to attract viewers at that time slot with his burned-out Tom Snyder routine.
I see he hasn’t posted for five days. I didn’t know he was fired. NYT probably owns his website too. Bummer. I used to read him every day, but when NYT put a limit of three times a month without a subscription, I pretty much stopped.
I hope he got enough to retire on from NYT, or was it just salary?
He wasn’t fired; from all indications, he chose to go to ESPN. The NYT threw a little bit of a shitfit in their public comments about it, insinuating that Silver was selling out for money. My guess is that money may have been some incentive, but sports is his first passion…so why not go back to what he really likes doing. He started political stuff with an online alias named after a pepper at Daily Kos and became quite a bit more famous/well-known than Markos could ever dream of becoming…not bad for something that he probably considered a hobby at best.
Thank you. I couldn’t find anything googling.
Markos who?
he was Poblana – never knew that
Poblano
http://www.dailykos.com/blog/poblano
I think the role of Nate Silver will be played in 2016 by Nate Silver, at his blog fivethirtyeight.com, which now reverts back to him.
I’ll add that his tenure with the Times never affected me one bit. I went to fivethirtyeight.com when he was independent in 2008, went to the same address when he was with the Times in 2012, and will continue to go that same address now. As a consumer, I never got any nearer to bothering with the Grey Lady because of his presence there.
There’s not much point in him riding a print outfit into the ground.
Given his major passion in politics is national elections, his political season is much more limited than his sporting one. There wasn’t much point in going to his blog on a regular basis unless there was a major election in the air. Sports punditry gives him a full time gig – he can now outsource a lot of the grunt-work at 538.com whilst leveraging his existing brand to broaden his popular base. It’s hard to see what the NYT added to his work besides putting some of it behind a paywall.
Hmmm … Nate Silver and ESPN … hmmmm
I wonder how long until ESPN asks him to work up a statistical model on Tim Tebow with daily updates.
I wonder how long after the first such analysis they ask him to stop.
Difficult to do that with 1 person.
Nate Silver had an audience and a web site before he joined the NYT and will have one after he leaves. I will always continue to follow Nate not only for his predictions but his thoughtful discussions of the science and math.
I can’t cry for the NYT. After the Judith Miller crap and the lazy, dishonest non-apology apology they gave later about that, my only sentiment for them is negative.
Honestly, I have no idea why people think that aggregating polls is the next thing to panning for gold. Folks, it’s not rocket science. It prosaic, common, everyday polling work, and he didn’t even do the basic data collection ( I am very sympathetic to the Gallup guy when he said that – he was right about aggregaters). They need someone with a MS in political stats.