Welcome, Science Friday listeners! (I’m on today, 2:15-2:40pm Eastern with Flora Lichtman and Nate Silver.)
Update: That was fun. We got to get into the basics: how news media love to be contrary, the weirdest polls (hello Gallup!) get the most attention, and aggregating polls is really the best approach. If you missed it, you can listen here (NPR).
and the Meta-Margin…
It’s dinnertime and dark here in London, but am listening to you and Nate streaming on Boston NPR on my phone as I get ready to go to the supermarket. Love the 21st century.
Can i listen to it on Sirius?
try NPR Now 122
http://www.npr.org/everywhere/sirius/
Looks like we’re on our way to a third “Red Friday” in a row, with multiple good swing state polls for Romney:
VA: R+3 (Ras), shifts VA from R+1 to R+1.5
FL: R+5 (Ras), FL stays R+1
NH: R+1 (PPP), shifts NH from Tied to R+0.5
IA: R+1 (PPP), shift from O+3 to O+1.5
Still there’s some good news for the blue team:
NV: O+8 (Mellman), O+3 -> O+5
NC: O+3 (Grove), R+3 -> R+2
WI: O+2 (Ras), stays at O+2
NBC/WSJ/Marist had O+8 in Iowa today, Florida poll by JZAnalytics/Newsmax was Obama +4
OV, I considered the Marist Iowa as yesterday’s poll (I saw it last night). The FL JZA poll was basically a repeat of yesterday’s O+3 poll from the same firm (with overlapping polling dates).
Froggy, the JZA poll for Florida has different dates and is counted twice on Pollster. Is it counted twice on PEC? Also, I noticed that there is now a bunch of discrepancies about which polls are listed on RCP, Electoral-Vote, and Pollster. How is the site reconciling with the difference?
Billy, my understanding is that the model skips multiple polls from the same pollster if the start and end dates overlap (like the JZA FL polls). The polls that are used by PEC are those that are in the Pollster database (if that’s the right word).
Froggy, why report that stuff?
Tunnel vision.
PEC provides the best poll mix.
This aint no red Friday!
Hello?
Red? Wow. Rose-colored glasses. Go to 538, ignore his analysis, and look at the data.
Meta-margin didn’t act like it was a red Friday… Or are they not in yet?
RDT, a couple of shifts toward Romney haven’t made their way to the MM yet. By my reckoning, IA should be O+1.5 instead of O+3, and VA should be R+1.5 instead of R+1.
But by far the biggest effect on the MM yesterday was the late afternoon Fox OH O+3 poll, which moved OH from O+2 to O+3. Even when IA gets updated, this means that there will still be 271 EV that are at least O+3, which makes for a good MM lead for Obama. Of course just one state pulling back could cause a significant drop to the MM. The floor under that 271EV would be WI (10 EV, O+2) and IA (6 EV, O+1.5).
Talk about the moon?
“Numerati”. Heh. Break a leg. :)
lol @Sam Wong
Afraid I won’t be able to listen at work, but I’ll look for it tomorrow.
Great! Did you say 8 out of 9 ?
My dad misses Stevenson…
Great show, Sam! They should have you and Nate on every week until the election!
That was great! Nice job, Dr. “Wong”! (heh)
That’s how he pronounces his name. I’m sure Sam clarified as he does with every intevirw.
Tried to listen to you Sam but I couldn’t. I thought it was a live event?
I wonder in the MM, does my home state of Va, have a chance?
ISee, it was. It ended a few moments ago.
I tried the whole time to connect but kept getting the program hasn’t been produced yet. This was on SCI-FRI’s website. WTH?
Just listened on Science Friday. Good work Sam! I particularly enjoyed your point about models that take too many variables into account (like unemployment etc) and in the process just introduce more noise. I like the elegance of your model….and i’m super happy you are updating more often now!
I find it particularly weird that Nate Silver uses so many econometric variables in his model since he’s previously 1) Warned of the dangers of overfitting, and 2) Blasted models based on “fundamentals,” which have been very error-prone on average.
I concur with Xtal. Weekly commentary?
Will there be a transcript to read? I couldn’t access any of the live feeds, and the local NPR out of Dallas had “Way with Words” on during the last hour.
Thanks, everyone. Erm, that *is* how my name is pronounced. Hey, you learn something every day…
My old contracts professor pronounced hers the same way, so in my head I’ve always said your like that.
Of course hers was my least favorite 1L class, so the association is not all positive.
Great to hear you on science Friday this afternoon. I have a question for you. As we get closer to election day, when if ever, would you feel comfortable saying the predictions will hold true on Nov. 6 barring some major world event or big story like the 47% remarks. Is that what the Median EV estimator attempts to show?
Chris, transcripts and audio should be available later today here: http://www.npr.org/programs/talk-of-the-nation/
Thanks, Froggy.
Sir, do you rhyme your name with Bang or Kong or? As to the weekly commentary idea, it’s excellent. however, a speakers’ bureau [agency] could do a lot more for Sam seasonally. It’s possible he could make a brilliant fill in this week for a speakers’ bureau’s cancellatins… We’d miss out, of course, but the reduction to the federal deficit would benefit us all.
If desired, I’ll make a few calls…
Glad to learn how your name is said. I _suspected_ she’d asked… So, I culdnt hear what Nate of the President’s chances, though of course as a 538 refugee I’m more interested to know- Did you say 8 out of 9??
I’m happy with watching all the charts. This is a gift for Dad. He’s 85.
I believe he said 70-30. I said 9-1, or 90-10. I offered him odds. Gambling nerd humor.
Nate had it at 70/30 for O. Sam had it has O winning 9 to 1.
Great job Professor Wang as usual. I enjoyed listening. If you don’t get the podcast they will have the audio here soon : http://sciencefriday.com/segment/10/19/2012/making-sense-of-presidential-polls.html
Recording should be available ~6:00 PM EDT
http://www.npr.org/2012/10/19/163245520/making-sense-of-presidential-polls
Meta-Margin just shot up over half a point.
Must have been the interview.
I think the noon MM dropping back to 1.32 was a screwup. It looked like it was due to the Colorado projection sliding 2.5 pts for no apparent reason, as there was no new polling. So the MM has really only ticked up .16 since this morning.
…shot back up to where it was this morning.
If the Rasmussen outliers were discounted, we’d probably be much closer to O +3.0%. Give it time…
All three of those make perfect sense, but I prefer Sam’s explanation.
aren’t outliers by definition discounted if you are working from medians?
Outliers still have an effect with a median, by virtue of existing; it just doesn’t matter how big they are.
If the data are +1, +2, and -9, then the average is -2, but the median is the middle value, +1.
No, Sam, you’re the rooster. And while you may not have moved the polls, you did wake us up.
Behold, what margin through yonder tracking breaks? It is the east, and Biden is the sun. Or Apollo, or something.
Olav: don’t worry about Ras. The median automatically downweights outliers.
Fantastic!
Just one advice for the President though…
Why doesn’t he use Hillary Clinton as a speaker? She will certainly get the woman’s vote out!
Secretaries of State traditionally do not campaign heavily. After the President they are the second-most visible face that we show to the world, thus they try to be as apolitical as possible given their particular circumstances.
It would be a violation of the Hatch act for her to campaign for anyone. It’s the one big negative of having her as SoS.
Guys,
Wait for 5PM. The 3PM didn’t bring the PPP polls in.
PPP is incompetent.
So it’s another tick down. I think I’m past fretting about these as the crack of doom.
PPP has variation, like any polls.
Yes, the interview! Put it, VP and D2 on the EV estimator :D
I was so excited to be polled just now– I got called by an NBC/WSJ (I think he said) poll. And then the call was dropped.
AT&T is clearly creating polling bias with its bad connections.
What state?
Jacob, I presume you are an outlier — relative to AT&T’s network, that is. Unfortunately that is something you have in common with very many Americans. ;)
I’m digging that MM climb.
Sam-
Nice job on NPR today and glad to be “introduced” to you and your website.
Great interview Sam. The interviewer gave you a nice compliment at the beginning saying you only missed the EVs by a point in ’08 and were dead on with the Popular Vote.
Great job Sam. I enjoyed listening to it.
For everyone else, has your life come down to the mm every few hours? Polls released today are razor thin. Make sure you vote.
And locate a nearby swing district for GOTV!
I just watched Chuck Todd and some guy from Politico stating that the race is clearly moving in Romney’s direction. Uhh…..wha?
Even Gallup ticked down from the R+7 yesterday, Rasmussen is back to a tie, so even the most mainstream of mainstream tracking polls shows a move back towards the President.
And that’s not even touching on the state polls.
Yeah, I know. It’s like the only poll they’re paying attention to is the Gallup tracking poll!
The same folks who’s profit goes up with the closeness of the race. Imagine that
Well, bare in mind that some of these people will do anything to make everyone think they’re worth the advertising dollars, ad clicks, &ct. I saw fox news spin Romney’s 38% of the Latino evangelical vote as “Romney leads with Latino evangelicals!” despite the fact the president is crushing him in every Latino category including that one.
Forget it, Reason, it’s Electiontown.
I am assuming that you were referring to the Hardball segment.
I am a long-time fan of Chris and occasionally email the producer. I attempted to educate them on scientific methodology vs. educated speculation (aka bloviating).
Maybe by the next election we could have a more informed discussion from the ‘gang of 500’…
What is with Intrade? Obama’s taking a beating the last couple of days.Not in line with 538 or Dr. Wang.
Intrade is a conventional-wisdom-meter. The conventional wisdom is very, very confused at the moment.
I think the polls will be volatile until election day. At least the National Polls. The NH and IA polls from PPP are a little bit troubling, as these are state polls. But then other polls have O ahead. I just cannot get past Todd and the other bums of the MSM making that comment today. But then, Sam did mention that they will probably pull that to keep the ad dollars coming in.
Seems like we had a broad range in the two recent IA polls. You have NBC/WSJ/Marist saying O+8, and then the PPP R+1. The aggregation of the polls seem to suggest Obama has a pretty comfortable lead there, but no matter how much I try to focus on the “big picture”, even outliers can make be uncomfortable.
Also, I wanted to ask Dr. Wang about why he moved NC back into the +R category. I did not see any new polls out today.
Well,
this pollster has him up by +3 in NC
http://www.projectnewamerica.com/latest/obama-leads-3-north-carolina
Thanks, Isee. I saw that earlier. Thus why I ask even again, how is NC back in +R category.
There are three NC polls in the last week on Pollster.com:
O+3
R+6
R+2
The third (oldest one) is the median.
Starting to look a lot like the S&P 500 at different time scale. Long slow bull followed by precipitous bear.
How is the S&P like the election exactly?
Very interesting data analysis Dr. Wang, I enjoy this site. I was curious, what is your take on Allan Lichtman’s 13 Keys?
I cannot make myself believe that NHis tied.I really believeObama wins there.
NH swung Democratic in 2004 and 2008, and that’s probably the long-term trend. But it’s still a purple state with large areas of libertarian-flavored conservatism. Republicans won huge there in the 2010 Tea Party wave.
It occurs to me that the real-estate bust and the subsequent re-consolidation of tech industry in the core of the Boston area might have halted the growth of southern NH as an effective Boston suburb, which might reverse the trend toward NH becoming more and more liberal as well. But we’re probably seeing shorter-term effects here.
I’m curious if anyone has insights as to why VA is as much a swing state as it is now? I was assuming it would be as red as red could be. Thanks.
A combination of minority voters and the out and out dissing of the 47% by Romney getting good traction?
Relatively liberal DC suburbs that resemble the northeast and have a whole bunch of people.
Northern Virginia is basically part of Washington DC. Heavily Democratic.
Then there’s the college towns (Virginia Tech, UVA…) then there’s the cities with large black populations (Richmond is 52% AA, Newport News is 41%.)
All of this counter-balances the rest of Virginia (mostly rural) which is pretty red.
Obama won in ’08 by 6.3% points (outperforming RCP poll average of 4.4% pts)
Note: PPP nailed it (final poll +6)
Actually, living in Va, you can now pretty much cut it not North to South, but East to West. And that diving line is Route 15. Everything east of it: major cities (including Richmond), Va Beach, NoVa, Fredericksburg, etc., is mostly Democratic. West is heavily Republican (except Roanoke and the college towns of Blacksburg, Charlottesville and Radford. Kaine is polling well here, so I am not sure why Obama is not doing as well in post Debate #1. So again, I am somewhat concerned about the polling trends in this state. I know it will be closer than 2008, but now with all this news of more election fraud being committed by the Republicans, concerns abound.
This used to be a daily stop, now I have to come by at least 4x a day. keep it up! Amusing to have MSM talking as if Romney is ahead, but makes you wonder about the quality of the rest of their news.
I refresh PEC to find Obama recovering nicely (now at 2.28%). And then I check the RCP average and he has reclaimed his lead! By .1 percent. Which is a fragile lead to say the least… But the trend is great.
Could someone talk to me about CO? RCP and this have the largest gap.
Different poll set, I guess. RCP doesn’t have the YouGov poll or the Grove Insight polls, which both have Obama at +3, and which are listed on the Pollster chart.
I think they might be a bit imperfect with updates. Otherwise a fantastic resource.
I hope that we can keep the lying dogs out of the White House.
Nice job on Science Friday! Glad to hear you and Nate end on the importance of state polls.
Now that the median EV and meta-margin are both shooting up like rockets and there is great rejoicing, let me inject my bit of cold water:
The movement is too sudden. It was too sudden on the way down, and it is too sudden on the way up.
Maybe Obama and the Chicago crew orchestrated a dive in debate 1 knowing they could turn it around in the VP and debate 2. And in the meantime, Romney looks briefly viable, and money that would have flown downticket flows back to Romney.
I bet Aaron Sorkin already thought of that one.
ALath,
Why is that cold water?
The Presidential race is mostly over. (Then again, I thought that in July.) The MM doesn’t do that unless it’s headed several more points in the same direction.
I will be getting back to downticket soon.
Sam C: cold water in the sense of casting some shadow of doubt on the entire polling ecosystem.
If the first derivative of the plot is too steep for the mechanism you are observing, then maybe your measurements should be questioned.
My hunch is Obama has been at 300 – 310 EV all along. All this up down stuff may be like the twinkling of stars: measurement effects, rather than real changes in amplitude.
But as Sam W said on radio, polls are the best indicator we have. And with more frequent polling, some of the effect will be mitigated.
That is an interesting point about 300-310 EV. I was pondering the possible consequences of a <10% response rate, which is what Pew reported this year - far down from previous years. One issue is that previous methods of defining LV might not work so well. Strong partisans might say "heck yes!" when asked to answer a phone survey. This idea is hard to test unless polls happen to stray outside that band, in which case we might learn something.
I also note that what you are proposing suggests that the race is even more stable than is shown. That is not my intuition. This year's Meta-Margin has a standard deviation of 1.3%, remarkably small to begin with. You're basically suggesting a deeper value that is even smaller!
Amitabh, I just think the composition of the electorate has evolved faster than our ability to measure it.
If the polls are just measuring enthusiasm, then you would see roughly the same abrupt fractals.
Diabolically clever. Create a giant diversion upticket, distract the opposition from downticket. The old “pretend to have a broken wing” trick.
Doesn’t this seem like an unlikely level of political chess? The game feels more like checkers to me.
Amitabh, I’m glad to hear someone had the same conspiracy theory about downticket money flows as I did.
Is it really possible that they tanked to keep Crossroads GPS and the like invested in the top-ticket instead of marginal House races? That would take some incredible confidence.
A week ago, John Scalzi proposed essentially the same thing:
http://whatever.scalzi.com/2012/10/12/for-your-friday-political-discussion-consideration/
Both cases are implausible, just because they require that the Obama campaign assume an unreasonable amount of certainty about how things would shake out. I don’t think they’re that stupid, or that omniscient. Everything suggests that debate #1 was a tactical mistake instead of a strategic mistake.
My take is that the six weeks prior to the 1st debate produced a huge “Romney is such a loser” bubble which artificially inflated Obama’s standing. I think there’d have been a sizable drop after the first debate even if Obama had performed well as a whole lot of R-leaning voters came back to the fold.
Totally agreed.