Princeton Election Consortium

A first draft of electoral history. Since 2004

Nerds under attack!

October 29th, 2012, 10:00am by Sam Wang



Paul Krugman is calling out National Review Online for their attempted takedown of Nate Silver for biased methods and somehow cooking the books. Krugman writes:

This is, of course, reminiscent of the attack on the Bureau of Labor Statistics — not to mention the attacks on climate science and much more. On the right, apparently, there is no such thing as an objective calculation. Everything must have a political motive.

Now more commentators on the right, including Jay Cost (The Weekly Standard) and Jennifer Rubin (Washington Post), are getting in on the act. Wow, dogpile on the rabbit!

A popular approach to undermining technical knowledge is to throw mud, assert expertise, make picky points, and sow doubts among the less savvy. In this case, what’s the argument? The NRO writer, Josh Jordan, makes this core criticism:

When you weight a poll based on what you think of the pollster and the results and not based on what is actually inside the poll (party sampling, changes in favorability, job approval, etc), it can make for forecasts that mirror what you hope will happen rather than what’s most likely to happen.

Jordan sounds like many partisan polling enthusiasts – on both sides. However, his style of poll-dissection can very easily lead a person astray. The human mind has a large capacity for finding reasons to reject a piece of disagreeable evidence. I’ve written about this in the context of how people form false beliefs in politics (“Your Brain Lies To You,” NYT, June 27, 2008). Polling internals lend themselves very well to such “motivated reasoning.” It is always possible to find something not to like in a poll. This is why I discourage all of you from chewing over single polls.

Silver’s evaluations of pollster reliability are quantitative parameters. However, there isn’t full transparency about how he arrives at them and what he does with them. This leaves him open to attack.

Partly because of this risk, I have stayed with simpler rules such as

Combined with a probabilistic calculation, these rules guided our Meta-Analysis to the exact EV outcome in 2004. It missed by only 1 EV in 2008. Such simple methods are easy to make transparent. You (or Jay Cost, I guess) could download my code in an instant.

I have my own technical beefs with FiveThirtyEight (for example, see here, here, and here). I believe Silver doesn’t extract all the information and tends to add unnecessary factors, which leads to blurry probabilities and poor time resolution. However, his intuitions about the data are excellent and he is very concerned with getting things right. For purposes of popular consumption, he is a fine and honest nerd.

Jordan’s capacity for wishful thinking is apparent when he writes:

While it’s impossible to know how the late deciders will break, the historical trend has been for them to break for the challenger.

I sympathize with this, since I thought the same in 2004, and added a “turnout/undecideds” parameter. For this I received a well-deserved drubbing afterward. In fact, undecideds split about equally, as amply documented by Charles Franklin. I don’t add such parameters any more.

(However, if Jordan wants to implement his idea, he can do so easily by clicking the “With +2% for Romney” link, over in the right sidebar.)

Finally, I will state something obvious. None of this storm of criticism would be happening if “Ro-mentum” (Oct. 23) were real. In fact, Mitt Romney’s fortunes peaked around October 4-9. Since then, the race has moved back toward Obama by about 2.5 points. National polls* give the graph at left. (See update, below.)

And the Popular Vote Meta-Margin, which describes how much state polls would have to swing to generate a tie in Electoral College mechanisms, looks like this:

History of Popular Meta-Margin for Obama

The Meta-Margin may still be catching up with national polls. If so, it has a few tenths of a point to go before it stabilizes. Alternately, something different is happening in swing states. In either case, the overall picture is the same: a narrow Obama lead that is static – or perhaps widening. There is no evidence for Ro-mentum.

>>>

*Analyzed as previously described. To generate the graph above, Gallup/Rasmussen were excluded. I am not at all averse to using them, but they have large house effects and so would need to be analyzed using median-based statistics, which I did not apply above. Anyway, adding them back gives the same relative picture of rapid decline and bounceback, with the same shape, except that the entire graph is slid upward toward Romney by 1.0%.

Median of all national polls, day by dayUpdate: OK, here is the poll-median graph, including Gallup/Rasmussen, day by day. It is not a weighted median, but it’s close enough (adverse blogging conditions). Note that the drop on October 3rd is due to a Rasmussen poll, which demonstrates their house effect, about two points relative to other pollsters. The conclusion from this plot, as well as the plot above, is that the race has been at a standstill for the last two weeks.

Tags: 2012 Election · President

297 Comments so far ↓

  • Matt

    National Review is upset that black votes count more then 3/5 of white votes.

  • Khan

    The Hudson River has breached and half of Manhattan is under several feet of water. Hoboken in Jersey is also under water. Subways are flooding. Ground Zero is flooding. There are cars floating down Wall Street.

    This is getting really scary.

  • Blair Mac

    What are all of these trolls going to do on the 7th, if Zomney loses?

    • skmind

      They will be blaming Dr. Wang and Silver for helping the librul media suppress voter turnout (among Republicans only, though) by falsely building up Obama.

      For a day or two. Then they’ll move on to Benghazi Part 2, and the upcoming fiscal cliff that Obama will be driving us over, so maybe they’ll discuss impeachment?

    • orchidmantis

      Buying tinfoil. Invest now.

  • Ed Groome

    Dr. Wang,

    I note that after I and others commented about the worrisome 1.5% slide in the graph [if Rasm. and Gallup were to be included], that you adjusted the slide to 1.0% without explanation and then deleted my comment after it had been up for a while. An odd business, that, but your site, your business and who knows, perhaps comments simply disappear when the weather cuts up rough.

    Seems to be a lot of adjusting going on here, though, for such a purportedly automated enterprise. 2004 all over again? That said, I do hope that your process, adjusted or automatic, ends up being accurate in the end. You’ve clearly put a lot of effort into it

    And while this comment won’t make the cut, I’ve enjoyed my brief sojourn here and learned some important things from you about polling.

    • Sam Wang

      You have made two factual errors.

      There is no slide over time. The entire time course is about 1.0% different, *vertically*, when G/R are included.

      There are no adjustments to the Meta-Analysis, ever. I make custom plots for my blog posts.

  • Marco

    First:
    Patrick Henry? Oh, Brother…
    second:
    Dr. Wang, I know you like to play with Jersey’s vote and all that, but indicating the power of a Jersey vote as 5.7641E-7 seems a bit much even for a precise calculation.
    Let’s call it zero, shall we? :)

  • Bruce Wayne

    Sorry. I didn’t mean to double post my last comment.

    I would also like to add that Silver leaves himself vulnerable to these kind of attacks because he chooses not to be completely transparent with his procedures. I’ve been following 538 since 2008 and I’m huge fan of his work, but it always irks me that his model is essentially a black box. He tells you what his doing, but he doesn’t let you see under the hood.

    I’m guessing he wants to keep his model a trademarked secret. The drawback of course is that it also takes away from his credibility. You essentially have to take his word that he is being honest and hope that he will be as accurate as he has been in the past. However, if he is in fact leveraging his credibility to up-sell Obama, there is probably no way for us to know.

    As a fellow number-crunching scientist, I appreciate that Dr. Wang’s procedures are simple and transparent. I have hard time reconciling that Romney has a <10% of winning when the polls are this close, but the numbers are what they are.

    • Dave in Tulsa

      I don’t defend Silver, because he doesn’t need defending and can do so himself. But there is tremendous irony in rightward commentators bemoaning his supposed bias in one breath, and then crowing about results from the likes of Rasmussen in the next. As with all things political, it depends on whose ox is being gored, but the indisputable facts are that Ras breathlessly announced the day before the 2008 election that Ohio was tied 49-49, that McCain was up by more than a point in Florida, and that McCain had pulled within four points in Colorado and Virginia (I won’t bother highlighting his other big misses that year, like NV, WI and MI). The final results? Obama took Ohio by more than 4.5 points, FLA by nearly 3 points, CO by more than 9 and VA by more than 6. I’m no mathmetician but Ras was obviously way off in his battleground state polling, always in favor of the Republican candidate and always in favor of a narrative that suggested that the Republican candidate was gaining momentum. Now, will anyone on the right mention these indisputable facts? Of course not, but that is why a Ras poll showing Romney +2 in Ohio actually translates to Obama +2 in what is likely to happen. Which, of course, is what Sam is showing.

    • skmind

      It is a prediction game. Silver will get vilified by the right if he gets it wrong, and will be vilified till we find out if he got it right.

      He gets nothing for being right especially if others do too.

      But suppose he gets it right, and most other do not.

      Then, keeping the sauce secret, his model has cachet value.

      Besides, it is not as if Dr. Wang’s site is unknown. It may not be as popular as Silver’s, but people know of it, and know of the track record.

      So these rightwingnuts who vilify Silver today for secrecy should then praise Dr. Wang for a full disclosure.

      Do they?

      No.

      The reason is simple. They are not interested in Silver getting it right. They’ll vilify him till Nov 6, and will go away if he was right. Or vilify him more if he was wrong.

      It’s a prediction, people.

  • Alfred G. Cuzán

    Living in the Gulf Coast as I do, I appreciate the challenging blogging conditions, and I wish you well during and after the storm. Hopefully in a few days you will give us the rationale for having excluded Gallup and Rasmussen from the calculations. Stay safe.

  • David

    Is there a room of these Republican goons trolling all these sites?

    “Typical lib-tard site! Romney is obviously ahead and pulling away fast. All your pseudo-intellectual shuckin’ n jivin’ won’t cut it on Nov 6th. Soon your great Communist in Chief will be gone!

    I keep looking up for a sputtering prop plane to drop thousands of leaflets with this stuff on it.

  • Patrick Henry

    Typical lib-tard site! Romney is obviously ahead and pulling away fast. All your pseudo-intellectual shuckin’ n jivin’ won’t cut it on Nov 6th. Soon your great Communist in Chief will be gone!

  • Ashbel green

    Sam – have you ever run the 2000 election through your model? I ask because of the butterfly ballot costing Gore several thousand votes, wouldn’t the model show him winning Florida?

    • Andrew Ferguson

      if you can send us the state-level polling data for 2000, sure, we could totally apply the calculation.

    • John Sawyer

      This is my concern–polls that are as close as many are, may not accurately reflect what will happen to the votes once they’re cast.

  • Pat

    Sam, Thanks for adding the red-yellow strike zones for the meta margin! Landfall soon…

  • Bruce Wayne

    I think conservatives are just upset that not everyone is buying into their “Ro-mentum” hype.

    At the same time, I don’t think you can simply disregard Romney’s lead in the national polls and say it’s all about state polls. The discrepancy between the state level polls and national polls, simply don’t add up. I have yet to hear anyone give an adequate explanation of why that is.

    Here is how I see the next few days playing out:

    1) If the national polls are correct and swing state are still catching up, then we should see swing state polls breaking towards Romney. This would be somewhat equivalent your 2% boost to Romney scenario. If the boost is significant, Romney could tip enough swing states to win 270EV. Otherwise, we will end-up with a Bush-Gore like scenario where Obama wins the EV but loses the PV.

    2) If instead the state polls are correct and the national polls are off, the national polls should settle into a ~1% lead for Obama. Additionally, Obama should win with ~290 EVs.

    3) Some combination of 1) and 2) occurring.

    I think scenario 3 is more likely. I think the state polls will tighten, but with Obama still leading/winning in OH, WI and NV. Those 3 guarantee him at least 270 votes. The winner of the popular vote is anyone’s guess.

    • Jay Bryant

      The one explanation I’ve seen for the discrepancy between the national and state polls is that the deep south is very red, while the rest of the country is more balanced. Sitting here in the one blue spec (Austin) in the midst of very red central Texas, I can believe this idea.

    • Matt McIrvin

      Obama has had a pretty consistent structural advantage of somewhere between 1 and 2 points in the EV calculation relative to PV. That’s been the case through the whole campaign. It probably is because his opposition is disproportionately concentrated in very red states.

      Now, there seems to be a hint that Rasmussen and Gallup national trackers have actually pulled away from even the rest of the national polls over the past few days, showing a Romney gain that nobody else can detect. People assume that Rasmussen is being consciously tampered with, but it could just be noise; it’s a small effect.

    • Tractarian

      This is correct.

      I think there are four possibilities, really:

      1) National polls are right and state polls are wrong.
      2) State polls are right are national polls are wrong.
      3) Both national and state polls are right; meaning Obama has a built-in advantage in the EC.
      4) Both national and state polls are wrong; that is, the actual state of the race lies in between.

      Three out of the four possibilities (2, 3, and 4) lead to an Obama win. Three out of four roughly tracks Nate Silver’s projection (72.9% as of this moment).

  • E L

    “It’s difficult to get a man to understand something if his salary depends on him not understanding it.”
    —Upton Sinclair

  • Jay Bryant

    Speaking of attacks on nerds, I saw some nitwit on Google+ dismiss Mr. Silver because he’s “feminine” and Dr. Wang because his name isn’t “American.” Critical think skills seem to be in shorter supply than ever.

  • Bruce Wayne

    First let me say that the attacks on Nate Silver are completely absurd. I think conservatives are just upset that not everyone is buying into their “Ro-mentum” hype.

    At the same time, I don’t think you can simply disregard Romney’s lead in the national polls and say it’s all about state polls. The discrepancy between the state level polls and national polls, simply don’t add up. I have yet to hear anyone give an adequate explanation of why that is.

    Here is how I see the next few days playing out:

    1) If the national polls are correct and swing state are still catching up, then we should see swing state polls breaking towards Romney. This would be somewhat equivalent your 2% boost to Romney scenario. If the boost is significant, Romney could tip enough swing states to win 270EV. Otherwise, we will end-up with a Bush-Gore like scenario where Obama wins the EV but loses the PV.

    2) If instead the state polls are correct and the national polls are off, the national polls should settle into a ~1% lead for Obama. Additionally, Obama should win with ~290 EVs.

    3) Some combination of 1) and 2) occurring.

    I think scenario 3 is more likely. I think the state polls will tighten, but with Obama still leading/winning in OH, WI and NV. Those 3 guarantee him at least 270 votes. The winner of the popular vote is anyone’s guess.

    • Sam Wang

      My take on the national polls is that wherever they are (which is hard to tell), they seem stable. In this context, change in the Meta-Analysis over the last 1-2 weeks is likely to be movement toward some equilibrium. I do not see a reason why it would reverse. For this reason I think (1) will not happen.

      In regard to the popular vote, this is a vexing problem. The two national-poll graphs at the bottom of my post tell the story. Median-of-all-pollsters gives a <1% Romney lead, so for the moment it is quite plausible that he would get more votes. But note the gray zone, which is the 68% confidence interval. The answer is quite uncertain.

    • skmind

      “At the same time, I don’t think you can simply disregard Romney’s lead in the national polls and say it’s all about state polls.”

      Yes you can, if you are talking about the electoral college.

      The national polls may well show Romney winning the popular vote, but it is a a hollow victory.

      As to 1), the national polls are going the other way, they seem to be catching up to the states polls. Remember, not too long ago, Gallup showed Romney with a 7 point edge among LV. At that point the meta margin was around 1.4. Today it is hovering around 2.0

    • JamesInCA

      Could the answer lie in sparsely-polled or unpolled states that are heavily red, e.g. AR, AL etc?

      Perhaps there’s a pool of Romney voters in the national electorate that isn’t captured by state polls focused on the battleground states.

    • tibber

      This is what I’ve worried about. the trolls at 538 keep insisting that the state polls are lagging behind the national polls and they always catch up eventually, but I don’t know if that’s true or not.

  • Andrew

    Prof. Wang:

    Polling results for Nebraska Congressional District 2 are available now here.

    http://www.omaha.com/article/20121028/NEWS/710289929

    49-44 Romney among registered voters, 50-45 among likely voters in poll from 10/23-10/25.

    In their 9/17-9/20, it was tied at 44 among registered voters.

    Perhaps you can add this to your analysis?

  • wheelers cat

    Since I am the twitter advocate and pop culture fetishist at PEC (apparently) I’d like to point out that this viral video has 1.5 million views since yesterday.
    http://www.youtube.com/watch?v=6TiXUF9xbTo
    Will it make a difference?
    ;)

  • Critic

    (Leave real address please. -SW)

    I don’t doubt that “poll partisans” on both sides are deeply flawed. However, the attack on Silver seems reasonable, until he answers it. I think a lot of Jordan’s article is baloney. I have no idea why Silver weights a poll by X differently than a poll by Y. It’s opaque, but I’m happy to believe he’s trying to get it right as much as possible if that’s all it is.

    However, and I don’t study Silver in depth, this accusation by Jordan needs to be addressed by Silver:

    “But look at some of the weights applied to the individual polls in Silver’s model. The most current Public Policy Polling survey, released Saturday, has Obama up only one point, 49–48. That poll is given a weighting under Silver’s model of .95201. The PPP poll taken last weekend had Obama up five, 51–46. This poll is a week older but has a weighting of 1.15569.”

    I would be very interested to hear why the same polling firm has a recent poll weighted less than an non-recent poll. Maybe there is a good explanation for it, but I don’t know what it is.

    Also, to think that any individual, no matter how good his/her intentions, can run a site like Silver (with all the variables that he’s added), and rise completely above all personal biases, is folly.

    Finally, interestingly, the Democrats attacked Silver in 2010 nearly as much as the Republicans are now. Seems like bad news means the team that gets that bad news will rip into you, no matter who you are.

    • Ken Dogson

      Some Dems attacked the model. I can’t recall anyone attacking Silver himself, calling him effeminate and what not.

    • wheelers cat

      That is a falsehood.
      Nate was not attacked by liberals in 2010 in the same fashion that he is being attacked by conservatives in 2012.
      Its not just forecasting mathematical statistics that are being attacked in 2012– its Dept of Labor reports, global warming, the president’s birth certificate and the CBO.

      I’m mortally sick of the “both sides do it” meme.
      The truth is, ideology is asymmetrical. So is enthusiasm and political behavior.

    • Froggy

      Darned if I know why this is so, but heading over to 538 and checking out the weights for the various Ohio polls, what Jordan cites is not unusual.

      The 10/24 Gravis (O+1) has a weight of 1.00102, while the older 10/19 Gravis (Tied) has a weight of 1.16272. Heck even the 9/9 Gravis (O+4.1) has a higher weight (1.05544) than the latest poll, but the 9/2 and 8/27 Gravis polls have miniscule weights. I thought I had an idea about what these weights were supposed to be, but now I’m not sure.

    • Sam Wang

      I do not think these factors affect the outcome much. For that reason I wish he had avoided it. The proper way to address house effect is probably an additive offset, not a multiplicative factor, but that would trigger even worse blowback.

      The flavor of critical comments here is different: usually factchecking as to whether the median is calculated properly. Which usually you are the one to address!

    • wheelers cat

      Froggy, house effect and credibility are two separate metrics that Nate combines.
      confusing, yah.

    • grandpa john

      And he was right in 2008 and in 2010 despite the attacks.
      Within his site he explains his methodology for anyone who wishes to read it. if NYT is satisfied with his work then he can tell his illiterate critics to stuff themselves, he doesn’t owe them a thing and he certainly doesn’t need to descend to their levels of idiocy to make some explanation that they won’t understand or accept anyway
      As someone once said, “never explain your actions or words, your friends don’t need it, and your enemies won’t accept it anyway”

    • Joel

      My off-the-cuff guess would be sample sizes, but there’s a reason why I’ve turned more to PEC over 538, especially in the late stages of the election.

    • Matt McIrvin

      Silver gets a lot of criticism here for the complexity of his model. It has a huge number of adjustable parameters, which all individually seem reasonable, but there are so many of them that it’s hard to see how he knows they’re set correctly.

      The PEC approach at least has the advantage of simplicity, though it declines to adjust for a lot of possible sources of bias.

      The funny thing is that all this opaque fiddling on Silver’s part seems to mostly have the effect of making Romney’s calculated win percentage larger, by magnifying uncertainty. It’s effectively a hedge.

    • orchidmantis

      Silver actually does not need to answer the criticism, because he has a job. And because for every mole he whacks, ten more will rise up, half of them saying the same thing. (“Romentum!!!”)

      Jordan can make his own prediction, and Nov 7 we’ll see which one is right. If the right had any confidence in this (besides the crazy (but not thin) unskewed polls guy) they’d have their own aggregate model, rather than nitpicking everyone else’s to try and find one reason why the numbers don’t say what they know in their guts is true.

    • Commentor

      Silver gives recent polls a higher weight and older polls lesser weight as time goes on until the older polls eventually phase out. It makes logical sense to me.

      Silver is making judgment calls, and that opens him to criticism, but he makes judgment calls in an effort to improve his predictions, i.e. he adjusts his system if it does not accurately predict the results. He is accountable for his judgment calls. Although he did receive some criticism, Democrats certainly did not attack Silver like this. And many came to his defense. The curious thing about the criticism of Silver is that his judgment calls tend to go in Romney’s favor, i.e. he gives Romney a substantially greater chance of winning than most similar sites such as this one.

      This site removes the judgment calls and simply performs a statistical analysis of the data, creating a demonstrably more defensible prediction.

    • 538 Refugee

      Silver goes into great detail about the manipulations he does. Hence my username because I got tired of the overly complex detail. Think of it as a game of horse shoes. Do you need to know why a horse shoe landed where it did or do you simply need to know where it landed? The problem with the former is that we are humans and can’t replicate the exact motions each time. If we could there would be more 300 games in bowling. Analysis ends and over thinking begins at some point. I think Nate has reached that point for some of us. The bottom line is Nate HAS laid it all out at some point but the question is how much reading do you want to do?

  • Les Honig

    I have a question for Sam or others that is driving me nuts and I don’t know who to ask so forgive me for using this particular blog on Nate…I have been following the Reuters/Ipsos national tracking poll since it’s inception and they usually have shown anywhere from a 3 to 6 point difference between registered voters and likely voters. In the last few days it has shown Obama with a double digit registered vote lead but that has shrank both days a full 9 points to the likely voter screen of 3 and now 1 point…This seems very bizarre to me…Am I reading it correctly? Here is the link to their data today:
    http://www.ipsos-na.com/download/pr.aspx?id=12137

    • Les Honig

      I don’t know if I made it clear above. Their screen is 9 points away from Obama the past two days..seems enormous.

  • Reason

    For all the hard work and trying times Dr Wang and others do for this site, some good tips from Mitt. https://twitter.com/MittStormTips

  • Ken Dogson

    Silver is very feminine. Wang isn’t even an American name.

    Romney wins in a landslide.

    QED.

  • John F

    Not to nitpick but shouldn’t Iowa be O+1.5? (O+8, O+4, O+2, O+1, E, R+1)?

    • Froggy

      The latest poll is the Gravis, with a polling date of 10/24. The Marist O+8 has an average polling date 10/16, putting more than a week older than the latest poll, and thus out of the pool.

  • ChrisD

    Unsurprisingly, Rasmussen is included in all of PEC’s eight closest swing state medians. So how do they compare?

    NV: PEC O+2, Ras O+1
    IA: PEC O+1, Ras tie
    OH: O+2, Ras R+2
    CO: O+1, Ras R+4
    VA: tie, Ras R+2
    NH: O2.5, Ras R+2
    FL: R+1, Ras R+2
    NC: R+0.5, Ras R+6

    (Even so, I agree that they should be included.)

    • JamesInCA

      Those are all basically consistent with a Rasmussen house effect of about 2-3 points +R, which has been noted in a couple of recent articles.

  • Herbal Infusion Bagger

    “. And the main argument seems to be that Romney is winning independents (particularly when you look at Gallup and Rasmussen polls) and Democrats are over sampled in pretty much every poll”

    Correct me if I’m wrong, but aren’t Gallup and Rasmussen both robopolls with little cellphone coverage?

    I think the two issues that could lead to a surprise (either a stronger Obama showing or a stronger Romney showing than expected) are the assumptions behind the LV screens, and the differences in responses of robopolls vs. live interviewers, and landline-only versus polls that capture the cellphone-only population. I’m betting that landline-only robopolls like Rasmussen are significantly overstating Romney’s vote, but we’ll see on Nov 6th.

    “and there is no way Obama can win this year, since Republicans are fired up and Democrats are discouraged. ”

    One wonders how one can simultaneously believe that independents ranks have swelled through Tea Partiers ceasing to identify explicitly as GOP, and that said faux-independents are more fired up about voting for a GOP candidate. There’s cognitative dissonance somewhere, either in the minds of pundits or in said self-identified independent Tea Partiers.

  • tibber

    I’ve been coming here to read comments since it isn’t overrun with right wing trolls like 538, at least not yet anyway. I just saw something on Facebook trumpeting a new poll showing Romney ahead in Ohio so I’m sure it will be even worse today.

  • Michael S

    I wonder how much of this attack on Nate Silver is merely cover for explaining away election fraud.

  • Matt

    Perhaps Tea Partiers are identifying as Independent. That would explain both a high level of D voters relative to R and independents swinging toward Romney. I find that more likely than some kind of persistant “over sample” whatever that’s suppose to me. I mean how do you explain disparate pollsters consistently finding more D identifying voters, particularly when D’s are supposedly demoralized, unless more people are actually identifying as D.

  • Richard

    My brother is a bright, mathematically literate engineer, who is convinced that Romney will win in a landslide, which after I asked him quantify, he defined as at least 54% of the popular vote and 350 electoral votes. He believes the polls are systematically biased toward Obama and will turn out completely wrong. His main source of analysis is Josh Jordan, Nate Silver’s NRO critic. And the main argument seems to be that Romney is winning independents (particularly when you look at Gallup and Rasmussen polls) and Democrats are over sampled in pretty much every poll and there is no way Obama can win this year, since Republicans are fired up and Democrats are discouraged. I’m paraphrasing. I think Romney supporters are also having cognitive dissonance in that they can’t believe Romney is tied or leading slightly in the national polls and yet decisively behind in the electoral college. Ultimately, we will know who was closer to getting it right, Nate Silver or Josh Jordan, come the aftermath of Nov 6. I really like that Dr. Wang has proposed multiple benchmarks for the different forecasting approaches. That type of post election assessment should prove very instructive and serve to inform future forecasting. Too bad there won’t be any benchmarks for the critics, except hopefully that they were flat out wrongheaded in their criticism.

    • orchidmantis

      Tea Partiers used to identify as Republicans (in 2008) and now identify as independents, but vote Republican. Thus explaining at least a part of the R loss, I gain, and Romney’s strong showing amongst moderates.

      Or there are a bunch of feral Republicans hiding in the shrubberies (all with proper voter ID) who have cleverly evaded pollsters all these months, but will emerge from their forest aeries to vote Republican next week. Then slip back into the wilderness until 2014 midterms.

      I believe someone did a study about smart educated people being smarter at finding data to support their existing point of view.

    • JamesInCA

      I suggest you ask him why he thinks Ds are “over sampled.” If he thinks it is because pollsters weight the reported outcomes by party ID, you should let him know that virtually none do (maybe Rasmussen?).

      Given that, how does he explain the mechanism by which Ds are “over sampled”? Do more Ds answer the phone than Rs? Why? And why more so this year than in past years, when the poll aggregates correctly predicted the outcome? An engineer confronted with a problem ought to be up to proposing a mechanism by which the observed behavior occurs.

    • Commentor

      What do Josh Jordan and others mean by the term “over sample?” Does that have any real meaning? Or is this another way of saying that they do not really believe the results from hundreds of polls regarding the numbers of voters who have identified themselves as democrats and passed the likely voter test?

      Question: how would a pollster adjust their polling methodology to meet the proposed “correct” sample for Republicans or Democrats? How would a pollster even determine what the “correct” sample is?

    • JamesInCA

      @Commentor: Intentional oversampling means is adding more of a certain category of respondent to your survey. For example, if you believe cell phone respondents have distinct views you want to be certain are reflected in your results, but you have too few in your initial set of responses, you might go back to the phones and call additional cell phone numbers until you reach some pre-determined target proportion. That would be oversampling cell phone respondents.

      What I suspect most people mean when they criticize polls for oversampling Republicans or Democrats is more along the lines of weighting, which is mathematically adjusting your reported results after tabulating responses. If, for example, you call 1000 people and ask who they’re voting for in the election, you might by chance end up with 600 women and 400 men in your sample. Since we know the population is more like 50% each, you would report a more accurate result by scaling down the womens’ responses, and scaling up the mens’ responses, “as if” each group had contributed 50% of the responses.

      Weighting is common and accepted with immutable factors whose proportions in the population we know accurately: age, gender, state of residence, etc. In contrast, party identification is not an immutable characteristic, but one that shifts; it is one of the things we want to measure.

  • P G Vaidya

    Dr. Wang,

    I have followed 538 for the last 5 years and have enjoyed the transparency of Nate Silver. I have been nitpicking in my mind about how he keeps the outliers and although the Monte Carlo simulations solve some of the problems, it is not entirely satisfactory.

    Therefore, it was a breath of fresh air for me to find your website, a month ago and I have accessed it several times a day, these days.

    I do have a question. The idea that the individual state probabilities are independent of each other is very debatable. There are, in a sense, various principle components that influence various states. So, if Virginia goes down the chance of North Carolina going down should increase, but not Nevada to the same extent. The joint probability distributions should enter your shortcut to the quadrillion calculations.

  • deckbose

    @mediaglyphic. Bobo is David Brooks. From Wikipedia: “The term “Bobo” comes from Brooks’s 2000 book, Bobos In Paradise: The New Upper Class and How They Got There. The word is a portmanteau of the words bourgeois and bohemian. In the book, Brooks (who self identifies as a “bobo”) argues that modern American culture is controlled largely by these upper-class baby boomers who mix counter-culture with consumerism. This sort of pop psychology is Brooks’s stock-in-trade.

  • Matt

    “First they ignore you, then they laugh at you, then they fight you, then you win.”
    Gandhi

  • R. Eustis

    Any stats guy who resorts to Looney Tunes clips as illustrations is, in my opinion, a badass.

    And the “dogpile” scene is a classic.

    • Matt McIrvin

      I would be very, very afraid of such an outcome, because it would probably provide the greatest available probability of a violent coup d’etat in the US. This is not a precedent we want to set.

    • orchidmantis

      The reason it isn’t being considered is that it’s nuts. The Republicans spent four years trying to make Obama a one-term president; they’re going to toss over control of the White House in exchange for the awesome power of Paul Ryan as veep?

      This does not give them *more* power than Romney in the WH, Ryan in the House as budget god.

  • TheVillageIdiot

    In the event of an electoral college tie, don’t rule out the possibility of a Black Swan event that gives us an Obama/Ryan administration:
    bit.ly/TQsLi8

  • Matt

    Woah, what’s up with South Carolina? Did I miss something?

    • ChrisD

      Changes in top line numbers from noon to 3pm:

      MM: 2.20 => 2.12

      NV: O+2 => O+2.5
      NH: O+2.5 => O+3
      CO: O+3 => O+0.5
      CT: O+9 => n/a
      SC: n/a => R+1.5

      I don’t understand how SC gets in the table. There are just 3 Pollster polls from the state, all done last winter: R+6,O+3,R+6.

    • Sam Wang

      That is a very good question. Investigating.

    • Sam Wang

      The answer is that HuffPost somehow dropped an old SC poll out of their regular update. For that state we were using polls (as opposed to 2008 results). We’re making sure that won’t happen again.

    • Matt

      makes me wonder if it was just a mistake

    • Billy

      I am guessing that there’s a miscoded poll in the feed?

  • Matt L

    The attack continues/escalates over at Politico and on Morning Joe. http://www.politico.com/blogs/media/2012/10/nate-silver-romney-clearly-could-still-win-147618.html

    • Matt

      He must be making a lot of conservative pundits nervous. If they really thought his methods were so flawed they wouldn’t pay any attention to him

    • Matt McIrvin

      I’ve noticed through this whole campaign that most people seem to think a stated 60% or 70% probability is a much surer thing than it actually is. You can see that in Joe Scarborough’s statement: any proposition that “could go either way” has to be 50%. (I wonder what he thinks of Sam’s predictions!)

  • Analytical

    I cant believe even Joe Scarborough called Nate an ideologue. I don’t know when these boneheaded left and right lean some basic maths!
    Joe was a congressman!!!

    • Bruce Wayne

      Scarborough is probably echoing what he hard on Limbaugh this morning. I’ve been following Silver’s blog since 2008 and have not seen comment/take a stance on any political issue. The most he has said is that he supports Obama. He’s all about using stats to make political predictions.

  • Olav Grinde

    Already a Democratic early voting edge in Florida!

    On Friday, the Republicans had a 5 % early voting lead over the Democrats. However, the Obama Campaign has encouraged Democrats to do in-person early voting, since this is more “robust”.

    How did that go?

    Well, in-person early voting only started this Saturday — and it started with a bang. Democrats have already bridged the gap! Here are the numbers:

    Democrats 42.0 %
    Republicans 41.5 %
    Ind./Others 16.5 %

    Relative to the total 2008 vote, more than 22 % of Floridians have voted early. That’s over 1.8 million.

    I would not be surprised if Democratic GOTV efforts widen their margin. We don’t know how Independent and Third-Party registered citizens will vote, but it seems safe to say that Florida is not beyond reach.

    http://elections.gmu.edu/early_vote_2012.html

    • 4BOinKY

      Looks as though the GOP has an edge in early voting in Colorado. Thoughts?

    • Pat

      @Olav
      Let’s hope the margin widens fast enough, because in 2008 the early voting breakdown in Florida was Dem 45.6% , Rep 37.3%. Can’t wait!

    • Ben

      @4BOinKY

      From the latest figures I’ve seen, republicans are up about 3% in colorado so far, at the same point in 2008 republicans were up 0.5%, democrats did far better in the second week of voting so we’ll have to wait and see.

      Still a shift of 2.5% towards the republicans in early votes is not enough to overcome their 9% loss in 2008 , they’ll still need to pick up a large number of independents to take the state

  • orchidmantis

    Can anyone explain step 2 in the following?

    1) Discredit Nate Silver’s model.
    2) ???
    3) Romney wins!!!

    Even assuming partisan underwear-stealing gnomes, I’m at a loss.

    • Resigned

      In this case, step 2 is profit. As is step 4. Step 3 doesn’t matter.

      These people attack Silver (and PEC, if they ever find it) because they are PAID to attack him, because there are literally millions of people who will willing buy, read, and listen to them attack him.

      It doesn’t matter what the actual outcome is. Romney winning would be a psychological bonus, but its probably a costly one since they won’t have that sweet right wing anger dollar to milk once an R is back in the White House.

    • wheelers cat

      Nate said it better on twitter.
      https://twitter.com/fivethirtyeight/status/262077837564076032

      Unskewedpolls argument: Nate Silver seems kinda gay + ??? = Romney landslide!

      srsly, we should all tweet. maybe we can get PEC to trend.

    • Justin S

      Step 2 would be discourage the Obama vote by saying over and over how Romney has “Mittmentum” and 538 and PEC are all wrong. Then it becomes a self fulfilling prophecy.

      Ain’t happening. We are voting and we know the models are correct. The whole smear campaign stinks of desperation. Can’t wait til the 6th!!

    • Wanderingew

      Step two is “create a self-fulfilling prophesy. I liken it to a dog chasing his tail and expecting to eventually bite another dog

    • Jack

      Step 2 is manipulate the vote, through voter suppression or worse, tampering with the electronic polls.

      Over the past three months, Republicans have been crying that Main Stream Media polls are skewed liberal. If the vote–after manipulation–shows that Romney wins, then they crow that their models are more accurate than the statisticians and scientists.

      If media call the election for Obama, they cry conspiracy and demand a recount.

      Either way, the voters are invalidated.

  • E L

    If Dr. Wang and his fellow Nerds are accurate about this election, and I’m sure they are, and again 2016, will their accurate prediction history discourage voters from voting because, to use Dr. Wang’s terms, voters will assume the results are “baked in the cake” long before the elections are actually held? Why vote if it’s a done deal?

    • Alan Cobo-Lewis

      If I knew you were voting I’d've baked a cake
      Baked a cake
      Baked a cake
      If I knew you were voting I’d've baked a cake
      Whack a do. Whack a do. Whack a do.

    • Resigned

      Why vote at all when we’ve established long, long, long ago that P=0 and C>0?

      The model only works if you assume that people still vote. If they stop voting, then the model doesn’t work, meaning they would have a reason to vote since the cake didn’t get baked. So then they start voting again, which means the model works again, which means…

    • orchidmantis

      I don’t think so. I certainly understand the polls are dependent on people actually voting. (And you would need an asymmetrical effect to affect the election.)

      What will happen is that, okay, pff, they called Obama ’08 and ’12 when everyone knew that was happening, and ’10 House shift, but (ignoring all the details from those predictions), they have missed all the reasons this election is different. For example, the fact that undecideds break for the challenger.

    • Shawn Huckaby

      @resigned- That reminds me of a recent fun bit of circular logic: https://plus.google.com/+RaymondJohnson/posts/CSXeyftovTJ

  • Olav Grinde

    “That means that an outright Romney victory would come at the result of systemic polling failure…”

    Or systemic cheating.

  • Geo

    Although conservatives attack Silver, you get the same qualitative result looking at the Rasmussen Report polls, at least up today — Obama with a narrow electoral advantage. If you gave Obama all the states he either led or tied Romney in Rasmussen polls, he would have consistently lead 277-261 over the last 2 weeks. The latest Rasmussen poll shows Romney leading in OH, which changes this. HOWEVER, there are some obvious questions to ask about the Rasmussen poll. First, from 10/23 to now the Ramussen Swing state poll had Romney with a 4-6 point lead. Yet, if you looked at Rasmussen’s polling for all the 11 states that make up their swing state poll, Romney trails in PA (-5), NV(-2)and MI (-7), is tied in OH, IA, and WI, and is up in FL (+2), VA(+3), CO (+4), NC (+6), NH (+3). How do you get a Swing State average of +4-+6 Romney out of a weighted average of -7, -5, -2, 0, 0, 0, +2, +3, +3, +4, +6? There is an obvious inconsistency here. How can the average be the highest order statistics in your sample? Weighting by likely voter population, would make the difference between individual state and the swing state poll LARGER. There is a similar, but less dramatic discrepancy in the Rasmussen national and state polling. It still shows Obama with a stronger electoral college edge than shown in national polling. Much attention will be given to Rasmussen’s latest showing Romney +2 in OH, but what Dr. Wang’s simulations do for us, is to de-emphasize outliers. The press love outliers, because they are large, new (it isn’t news it if isn’t new), and generate excitement and readership. However, if your purpose is to predict rather than generate excitement, outliers should receive less weight.

    • Obama 2012

      from my perspective this is clearly a last ditch effort by Rasmussen at narrative setting… they know they can get away with being awfully wrong and keeping their reputation (they were way off the mark in 2008 & 2010… and yet here they are still being taken seriously) so they are willing to be wrong on Ohio if it helps create a story that helps their party. …

    • mediaglyphic

      Wheelers cat,
      who is BOBO?

    • Brett Wilkinson

      Bravo. Well said. In Ohio, we can’t wait for this all to be over.

    • wheelers cat

      Bobo is David Brooks.
      Chunky Bobo is Ross Douthat.

    • john b

      @mediaglyphic:
      BOBO generally refers to david brooks of the new york times (and NPR and general moderate right bloviating)

  • skmind

    This tactic has been used for years, and it is working.

    It starts off by a stance that glorifies scientists, none alive, but someone who is well known. Newton, Einstein (but never Darwin).

    Then it sets the bar to be very high. In other words if you, Dr. Wang, predicted the 2008 electoral vote exactly, then, you were doing science. But it is not even acknowledged as such. A jocular gibe at “even the blind squirrel gets the acorn” is thrown in.

    Scientists are too laid back and too busy to engage in this. So when they get it right, which is most often, it is ho -hum stuff

    But imagine if there was an error. That is amplified as an indictment of science itself, and the utter criminality of the scientist involved.

    So from their perspective, they hold scientists to an error rate of 0, while they pay no penalties for being 100% wrong, as long as they were right just once about some backwater election.

    Statistically, if you employ this strategy, over thousands of runs, you’ll be a winner.

    That is the danger.

    How to counter it? We need activists who will belabor the point that the NRO or whoever that rightwingnut was, was totally wrong, and a hack.

    And then repeat it.

    And then repeat it.

    Till, no one treats them as anything else but hacks.

    Until then, I’m afraid that they’ll win.

    So how best to go about it? Scientists do not have time to engage in this mudslinging, and they should not be expected to.

    Our mainstream media is either too focused on the next scandal, or is part of this mudslinging itself.

    Professor Krugman is an advocate, but he is careful to remain that and not be an activist.

    What we need are activists. Bloggers, people who go to city council meeting and yell about the abuse of science, promote each other’s blogs, highlight how the same NRO wingnut was wrong about so many other things.

    It has to become part of the mainstream consciousness.

    If things hold, and Obama wins, these idiots will go unpunished. That is why they do it again.

    And again

    • wheelers cat

      But they are being punished, skmind.
      The reason the NYT hired Nate and HuffPo hired Blumenthal is the rise of third culture intellectuals.
      The first culture intellectuals like Bobo are losing market share.
      Dead tree media is dying.
      There was a systemwide push back against Ro-mentum. That is epic.
      And now there is a pushback against the War on Objectivity.
      Its epic.

      Death rocks and evolution rolls.
      ;)

    • skmind

      The PuffHo (corrected your misspelling) hiring Blumenthal is the kind of reward I am talking about.

      Think about it, Blumenthal could earn more by parroting the NRO, and he would not even have to do “math”

      Now he is relegated to being viewed as a left-wing shill, and competing with the “Sideboob” section (yes, PuffHo has a WHOLE section on it) for his reputation.

      Nate is slightly better. He only has Brooks and Douchehat to taint him. And Krugman is on his side.

      Maybe the really sad thing about evolution is that it takes too friggin’ long. And therein lies the irony: conservatives who think it is a hoax don’t like it because it takes too long for a change to be perceived.

    • Resigned

      @Wheeler’s cat

      Oh how I wish that could be true. The sad thing is that it seems that the nerds can only ever lose. If they’re right, they’re right, and if they’re wrong (or not even “wrong”, merely “not 100% accurate about everything ever”) then they’re just a bunch of overeducated hacks.

      As for Bobo, he’ll still get his giant checks for spouting 800 words of nonsense twice a week, no matter what happens. The folks at NRO will always have a paycheck waiting for them. Newt Gingrich will always get invited back on MTP, no matter how off his predictions are.

      Sigh.

    • Vicente

      @Resigned, no matter how bad the war on objectivity may look right now, I would take wheelers cat’s side if I were betting on the long run. For an interesting parallel go onto espn.com and look at the boom in sports writers using advanced stats in Baseball, Basketball and now Football coverage – they’ve gone from a couple of writers in total to three or four per sport over the last few years. Most of them are behind a paywall that suggests people (like me) are willing to part with money to read analysis that’s actually grounded in data rather than anecdotes.

    • Billy

      As an example, scientists were jailed in Italy for not being able to predict an earthquake.

    • hs

      The trouble is that science doesn’t (well, at least, it shouldn’t) do activism. Science is based on the premise that we are doing what we do based on what the best understanding we have of the universe. We may not know everything, so we might be wrong. But, if we are wrong, we will learn from it. Activism (and punditry), on the other hand, is based on the belief that we know the answer, we are right, and if you don’t agree with us, you are not only wrong, but you are against us. If scientists have to resort to activist tactics, then it’s no longer a science because we’d be abandoning the possibility (essential to “science”) that we could be wrong and we should learn from it, if we are…

    • skmind

      @hs, no I specifically did not suggest scientists getting involved in anything beyond doing what they do best.

      The call for activism is to the rest of us to ensure that the wingnuts pay a price when they get it wrong. The problem today is the scientists get little credit when they are right, but a lot of negative publicity for the slightest of mistakes.

      The wingnuts are the opposite. They get no penalties for their reputation when they get it wrong (which is most of the time), and are revered as heroes for one correct scoop.

      That imbalance needs to be rectified. One way to do that is to collect these nay-sayings and then keep reminding wingnuts how wrong they were.

      Each time.

  • Joel

    @ vladimir

    Within the confidence interval, yes.

    That means that an outright Romney victory would come at the result of systemic polling failure (i.e. the null hypothesis that polling is meaningless could not be disproven). Someone can correct me if I’m wrong.

  • Obama 2012

    I listened to a little bit of right wing radio in the car on my way home today… it’s pretty amazing. They seem pretty much convinced that Romney is winning big. I don’t know whether it’s a sort of coordinated effort to create a self fulfilling prophecy or if they actually believe it themselves.

    • grandpa john

      I doubt that they really believe it, This is just a CYA approach so that when Obama wins , they can claim that Acorn of some other nebulous outfit that doesn’t exist,really stole the election.

    • WishfulThinking

      I’ve also seen almost non-stop coverage of Bengazhi by Fox News. Almost seems like an attempt at SwiftBoating.

    • Patrick

      I know some people do indeed believe it. I live in a liberal city, but my family lives in the ultra conservative suburbs: we’re talking one of the most republican areas in the country. My family members and their friends are completely convinced Obama will lose. They aren’t even entertaining the idea that the President might win; they are looking forward to next Tuesday as the end. There is going to be a SERIOUS meltdown if it turns out like I think it will. People there assume everyone in the country hates the president as much as they do, especially since they are insulated and surrounded by like minded middle aged, Christian White Folks.

    • Some Guy

      I think it is yet another instance of rightwing media doing anything they can possibly do to de-legitimize Obama’s presidency. So if he wins as the aggregation of data indicates is the more likely outcome, then they can say something was fishy and have a second term treated as illegitimate. A depressing portion of the country will not accept he is president rightfully and lawfully. What that is, I like many others, believe has to do with race, but also with the right’s careening, tottering lurch toward anti-democratic practices. Get the wrong answer from the electoral machine, then break it. I am genuinely worried about the “meltdown” that could occur. At this point, there are only villains to find in the event that Obama wins.

    • otis29

      Yeah, Hugh Hewitt not only says Romney’s going to win, but he’s going to win huge.

    • Some Body

      Some Guy – To be fair, many Democrats (and some of the regular commenters here) refused to accept Bush Jr. as a legitimate president, insisting that the 2000 election was stolen, and a smaller group insisting the 2004 election was stolen too.

      Now, that didn’t break the machine (Bush himself did much more damage). I wouldn’t be so grim about the prospects of what will happen if some Republicans end up feeling the same this time around.

    • Silvio Levy

      Some Body: There is nothing fair about equating a denunciation the Supreme Court’s decision preventing the State of Florida from following its own constitution (which mandates that votes should be counted according to the intention of the voter insofar as that can be discerned) with the hue-and-cry that we will hear when Obama wins a second term.

      Silvio Levy

    • Jack

      The prediction that Romney will win is part of their communications plan to discredit the election if they lose.

      First, they demean the polls in the Main Stream Media, saying they’re all biased.

      Next, they inject their partisan polls into the national averages to make the national polls closer, saying that their partisan polls were closer to “reality” than the MSM polls.

      Limited or suppressed turnout plays into the narrative by claiming voters who favored Obama in 2008 weren’t excited enough about the president to vote for him again.

      Finally, if Romney loses, they call “conspiracy!” They’ll call for recounts of key districts/states. They’ll try to eliminate voters (or change electronic votes) to match their partisan polls.

      I’m not a conspiracy theorists in any way, but this has been apparent to me since the Rassmussen, Gravis and WeAskAmerica polls started skewing dramatically about three months ago.

  • Paul

    Yeah, that article was disgusting. And I’ve seen worse.

    As long as we’re comparing PEC to 538, I’ll throw out this criticism of PEC: I think the fact that you’re not correcting for house effect makes your trendlines much jitterier. For small effects on the scale of days, PEC seems over-sensitive to which polling organizations happen to be the most recent. Nate Silver’s knife juggling really does seems to help smooth this out, and make a new Rasmussen poll more comparable to an old PPP one.

    However, I really do appreciate having PEC’s dead-simple approach out there, just saying, “This is what the polls say if you take them completely at face value.” That really pays off today as you leap to Silver’s defense.

    It lets you make a point I really wish you’d driven home more explicitly in today’s article: all the weighting and adjusting of polls has often made 538′s prediction *more* favorable to Romney. PEC’s method demonstrates that if that NRO bozo gets his wish and you look only at “what is actually inside the poll,” things look even worse for Romney.

  • Reason

    Nice article. Is Dr. Wang still doing okay up there?

    • Sam Wang

      Lots of rainy day activities here.

    • scott

      There seems to be vote rigging in presidential elections, including
      2004 and 2008, that can be seen by looking at statistics.
      Fraudulently, a computer program (possibly at the state’s central
      precinct tabulator level) automatically flips a percentage of votes
      from one candidate to another but that the percentage of votes flipped
      varies with the size of the precinct. The “rules” of the fraudulent
      vote flipping algorithm are the following:
      Very small precincts don’t have any votes “flipped” unless that
      precinct is larger than, for example, 350 voters.
      The percentage of votes that are flipped is small (such as .01%) for
      small precincts and large (such as 5%) for large precincts with a
      *gradual* change in percentage “flipped” dependent on the precinct
      voter population.
      The reason that the perpetrators don’t flip as many votes in small
      districts is because a recount will check a random number of precincts
      and a smaller precinct is *more* likely to be audited because there
      are more of them. So if fraudulent vote flipping flips 5% of votes
      from Democrats to Republicans, a random recount of precincts would
      show a smaller error – such as 2%.
      There is a smooth gradual change in the percentage of votes flipped
      between the small precincts and large precincts and is seen in the
      data.
      The authors of the paper show that the effect does not happen in data
      from some counties presumably because the perpetrators did not have
      access to those tabulating computers and it is not seen in democratic
      primaries. The authors of the study look at income and poverty rates
      which are highly correlated with voter choice but do not correlate
      with precinct size. This rules out more Republicans living in large
      precincts as being the cause of the anomalous data.
      To find the article, do a google on the following words: vote
      flipping small large precincts central tabulator
      Also, here is a link to the article,

      http://www.themoneyparty.org/main/wp-content/uploads/2012/10/2008_2012_ElectionsResultsAnomaliesAndAnalysis_V1.51.pdf

    • Reason

      There he be. I just saw on the Weather Channel that the storm picked up speed and may beat the high tide and move through the area faster. Still got power, Doc?

    • Sam Wang

      If you can read this, I probably have power.

    • Reason

      @SWang Hah! That was why if you could still log on. Stay safe.

  • Bigby's Forceful Hand

    To the ramparts, nerds!! Quickly!!

  • Bill Costley

    This website is far from Rom-mania.

    • Liz

      I just wanted to mention that Joe Scarbough on Msnbc made a statement this morning that anybody that believes the nate silver polls is basically stupid. Of course he cherry picks what polls he wants to give attention to.

  • Slightly Skeptical

    Sam, if you’re able, could you please elaborate on this: “To generate today’s graph, Gallup/Rasmussen were excluded. This doesn’t mean I am averse to using them. It’s just that the method I devised needs a bit of work before it can be used with median-based statistics”?

    I looked at the link you gave, and I really like this idea of spreading out a poll’s results across the days over which it was conducted. But I don’t understand what you meant in this quote.

    As regards the bigger picture, an electoral college vs. popular vote split is (to me) looking increasingly likely.

    Hope everyone in your neck of the woods is manages ok with any flooding! And sorry if this shows up multiple times, I tried to post it before and it didn’t show up.

    • Sam Wang

      It means if I averaged them in, they would have a disproportionate effect. My preferred approach is to use medians. However, this is slightly more complex when I also have to spread a poll’s respondents over multiple days. The patch was to calculate an average, but leave them out.

      See the update, which is a median of *all* pollsters. It’s not exactly weighted properly – will do that on a day when I am not lashing lawn furniture down.

    • Slightly Skeptical

      Thanks, I get it now. Before, when you had written ‘average’ I assumed you meant median rather than mean, since you tend to use medians in general. But now I understand. And sorry again for the multiple posts. I hadn’t enabled cookies for this site, so I didn’t see that it had gone into moderation.