The party conventions begin
The Republican Party’s national convention starts today [schedule]. Conventions are a chance for a political party to showcase their unity, the...
58 Comments
Leave a Reply Cancel reply
Related Content
Politics and Polls: a podcast with Julian Zelizer
It’s a new collaboration with Julian Zelizer over in History. The first episode, produced by the Woodrow Wilson school, is posted here, as well...
Slow news month ahead?
I’m on a low-posting regimen for a little while. Basically, I think the Democratic and GOP primaries are settled. That said, why don’t you...
https://i.imgur.com/UHAhKtZ.jpg
I don’t know.
This was so great! I was jumping up and down to see you on MSNBC. The Beyonce split screen was just gravy.
Sam, you are destiny’s child.
So is she a Bey-sian or a freakuentist?
I bet she is still be so excited about this!!
Polling just in shows that >99% of the lady bugs in the Princeton dorms as well as the house spiders and other insects on Princeton’s campus are aware of Sam’s bet and have absolutely no fear that any of them are at risk of being eaten!
To follow up, an empirical study is scheduled for this coming Tuesday to see if those bugs continue to freely cavort over the coming days, or if they go into hiding as Tuesday evening approaches.
Almost all polling places are withing a few blocks. Not on the other side of town.
I am getting worried about the Philadelphia transit strike.
Almost all the Blue votes in PA come from the Philadelphia and Pittsburgh. Anything that hurts turn out there would be very difficult. Add to it, Trump has been talking about Phila stealing the election. His mob connections. Mob connections with trade unions. Or some faction of the trade union being sympathetic to Trump. I am getting really worried about it.
Looks like no one else is. So not sure how big a deal is this.
I thought I heard they are not striking on election day. Plus, most people just walk to where they can vote.
3. Most voters live within walking distance of their polling place. It’s not known exactly how many voters would rely on SEPTA to get to their precinct but that number is thought to be small.
The management hoped and asked for assurance from the union. It declined. They sued. The court refused to force the union to work. Hearing on Monday.
Aw come on! Here’s Sam trying to distract with Beyonce and a funky tie and you’re just ignoring it and going ahead and freaking out anyway. Stop it.
So then they will walk to their polling place. Have you ever been to Philly and see how polling stations are set up for the city?
Good thing there are approximately a million polling places in Philly and everyone can walk to theirs…
@Amitabh Lath // Nov 5, 2016 at 3:27 pm Come on, how can one be a true blue dyed in the wool Democrat if one does not worry about Democrats somehow fumbling it?
Everyone can walk to the polling place, good. Hope they can get back from work in time to get in line before poll closing.
Its not a problem getting to the polls, the challenge is that people commuting to work within the city or those who work there and vote at home in the suburbs have a longer commute due to the strike so may not make it to the polls by 8 pm especially those with kids to pick up. I’m counting on the Septa Union to do the right thing.
So, if public transit isn’t running and some people are working from home, they’ ll have more ability to vote. Others will wind up figuring out a car pool, and they will have time to talk with one another and reinforce sentiments fueling their votes.
In short, the “possible” elements can be spun out either way. This site attracts on the basis of data – any chance someone has actual historical data of an impact of a transit strike on elections?
Uber & Lyft are offering free rides on election day in Philadelphia. Strike continues as talks resume
http://6abc.com/traffic/uber-lyft-zipcar-offer-free-rides-on-election-day/1592582/
Awesome hair. Beyonce looks good too.
I watched Smerconish this morning. I think I might be becoming a Sam Wang groupie. Is there a support group for that?
How about if we wait a week, and then I can return to my ivory tower! We’ll always have this moment, though.
So, Sam, what’s Beyonce really like?
Not much will impress your undergrads more than when you casually let refer to “that time I was on television with Beyonce.”
Man, Silver is really getting aggressively testy about complaints about his model today.
Watch the whole episode. Beyonce all the way. Maybe the best marketing strategy: data science is sexy.
Senate! I’m phone banking for Wisconsin Feingold!
In terms of marketing I’ve read there is more to these concerts than meet the eye. They are also registration drives and often take place in close proximity to polling stations or the place people pick up tickets are close to polling stations.
http://www.politico.com/story/2016/11/hillary-clinton-concerts-strategy-230784
Nothing sexier than data. I’ve been waiting a long time for the media to pick up on this fact.
https://twitter.com/SamWangPhD/status/795007244756799489
To all those insisting: “99%! It’s the facts. It’s the numbers. It’s science.”
… The facts are under consideration for adjustment.
[Note: I buy the 91-93% confidence btw. I just think that Russian Roulette w/ a 100 chamber pistol is a lot different than one w/ 11 chambers.]
I figure “>99%” is probably high, but it’s probably not a sensible time for Sam to go tweaking the parameters of his model. Might be worth taking a look at the assumptions after there are real results on Tuesday–to my mind, having a sharply defined hypothesis to test is more important than being right at any given time.
Well, yeah, there is *no* way I am tweaking anything. I am just engaging in some chinstroking about how far off “>99%” could possibly be.
I’m genuinely confused. Is he saying his model has a problem, or not? 99% or 91-93%? I agree with JSchmoe that there is a HUGE difference between those scenarios. I didn’t really understand Sam’s comment on Lawrence O’Donnell to the effect that any estimate over 90% is pretty much the same. Perhaps the urgency and uncertainty of the Senate races makes the difference between 91% and 99% seem sort of silly, in his mind. I can see that.
I’m also kind of confused. Wasn’t the bug-eating pledge tied to the model hitting 95%? If it were down at 91% would there be no bug pledge at all? Also, would shed new light on the Silver-Grim spat, as HuffPo would be somewhat alone on its own limb.
I’m sort of disappointed Dr. Wang was not wearing a pants suit that matched the background dancers.
Actually, my pants did match. You just can’t see them.
Karyn and I enjoyed watching your segment. I hope you were pleased that you got to describe your methodology AND talk about the senate races.
It may be a low bar but this photo is hands-down the highpoint of this election cycle.
Beyoncé aside, how cool is “Breaking News: Sam Wang, Princeton University Data Scientist”?
Nevada and NH going blue again looked reassuring, but it’s Ipsos polls whose survey periods go so far back, they’re subtracting information.
Eh. You’ve also got polls in there that were conducted in a single day, and polls that span a week starting the third week of October. It’s just the nature of public polling: you’re gonna get a lot of junk. Somehow, the midpoint of all the junk is usually not junky.
That makes sense. Thanks
@Josh // Nov 5, 2016 at 6:49 pm:
” Somehow, the midpoint of all the junk is usually not junky.”
A number of people guessed the weight of a cow in a UK country fair. The statistician noted all the guesses and published a paper showing the “average” guess came within 1% error. This was surprisingly recent, just 1905. (The mortality tables and actuarial science stretches back to Adam Smith and beyond).
Recent re-reading of the Francis Galton paper shows that it was the median, not mean that came within 1% of the correct answer.
So yes, Sam uses median and that is the right thing to do. “Mid point of all that junk is not junk” is probably the foundation of many modern systems.
I was re-reading 2012’s “A final unskewing” post () and noted that while “the bias in state polls turned out to be close to zero…national polls were biased by 2.4 +/- 0.4% toward Mitt Romney.” If a similar bias were in effect today, it would help explain why the electoral map looks better for Clinton right now than the bulk of the national polls seem to imply.
Not sure what accounts for this bias, if it exists.
Actually I don’t see too much difference between the state and national polls at the moment. The Meta-Margin is Clinton +2.7, and that’s really not too far off from her lead in the most recent national aggregates I’ve seen.
(I guess HuffPollster’s, at least if you choose “less smoothing” to get the recent skinny, is dominated by polls like IBD/TIPP that do drag it down a bit–I haven’t done a median calculation.)
The argument against taking only the most recent subset of data is that you can convince yourself that there’s movement when really it’s just noise. In an electorate as polarized and stable as this one, I’m convinced that’s what’s happening now.
If you look at only the last 7-10 days of national polls it looks like Clinton has a lead of around 3%. Zoom out farther (like 2-3 weeks), and it’s more like 5-6%. This, IMHO, is closer to her actual current lead. Also–and I think not coincidentally–this is not far from what her national lead would look like if her Meta-Margin was up above 4, which is the average of what it’s been all campaign.
On the 1980 polling “miss,” more evidence says it wasn’t a miss at all. Like I suspected, the fact the electorate was far more fluid also lead to many last minute changes after last poll. Carter had a very bad week after last poll, with Iranian hostages and inflation dominating the news cycle.
http://themonkeycage.org/2012/08/what-really-happened-in-the-1980-presidential-campaign/
Queen of pop and King of Stats. It’s a sitcom concept work in progress.
I made sure to tune in, having seen the post that Prof. Wang would be on the show. Good job guiding Lawrence through the stats…he always wants to put a little too much weight on favorable info, and I thought the explanation on what the 99% means was very helpful to the audience.
I agree Davey.
“Queen of Pop and King of Stats” – love it…
Many thanks for your efforts Sam!
Here is a link to the MSNBC clip:
http://www.msnbc.com/the-last-word/watch/scientist-predicts-99-chance-of-clinton-win-801634371744
Who is Beyoncé?
You’ve changed, bro.
No, we’re still good. Really. Hey, gotta go.
so when will we treated to a split screen of Dr Wang with Katie Perry and J-Lo?
Professor Wang, I have a question. According to your interview you use state polls only to compute your predictions. Would it make sense also to use early voting statistics in some way as they become available near the election? I’m thinking of the apparently significant Hispanic/Latino turnout in states like Nevada, Arizona, Florida and North Carolina, to name a few.
That sounds outside the purview of poll aggregators, but you might be interested in what these people are doing on Election Day:
http://www.slate.com/articles/news_and_politics/politics/2016/11/follow_slate_and_votecastr_for_real_time_election_day_turnout_tracking.html
I’m not sure it’s actually a good idea, but it’s potentially interesting. They’ll be using early voting data for their projections where available.
Sam, you make your acad. dad proud.
Sam
Great job on MSNBC.
I am so tired of people mis-using or misunderstanding statistical probabilities and what you are doing is presenting data and your model/hypothesis so simply that it hides how much energy you have put into developing and maintaining your model.
There is an old accounting joke – “Guy asks an accountant what is 2+2. The accountant responds “what do you want it to be.” That is how polling data appears to have evolved into and as a result if all polls are suspect -people will only trust the polls that align with their “gut.”
I loved 538 in the last cycle, but it has changed and Nate is making the classic statistics error of becoming a filter to the data as opposed to just analyzing the output.
Polls in early voting states that dont separate out those that have already votes from those that still need to vote on Tuesday are garbage. And dont get me started on sample size and sample population malpractice that show fluctuations outside of any reasonable standard deviation.
The media is making a killing this cycle with ad revenue and a hyper focus on a horse race that waa never as close as any news outlet would admit. It isnt conservative or liberal bias, its capitalism and unfortunately there are too many hack pollsters (notably not statisticians) that will give the media whatever they want 2+2 to be.
Beyonce and Sam, a “pair to draw to”.
At least Sam got the prestige spot – right hand side, portrait mode, close-up.
I had high hopes that the projected numbers would be close to reality, needless to say those hopes have gone up in smoke.
This makes me wonder how the polls missed so many voters who turned out for Trump. He said he had a hidden army of supporters and no one believed him, myself included. He was right and we never saw it because of all the other noise going on at the same time.
We’ll survive, we always do, if only because we have no other choice but to make sure that we survive.