Subscribe via RSS Feed

Silver

[ 229 ] October 29, 2012 |

The threat Nate Silver’s mathematical models offer to traditional punditry, with its emphasis on the horse race and the personalities and the media ratings, is quite real. How do we know? So many pundits hate Silver with the heat of a thousand suns.

And these are political pundits we are talking about here, people. Political pundits have absolutely no accountability to anyone. The idea that Silver could be discredited if he is wrong next week is hilarious coming from, say, David Brooks.

..[SL] If you don’t want to reward Politico’s trolling with your hard-earned links, Elle Reeve has a good roundup of more people who think Silver is a fraud because everyone knows that if a coin comes up heads twice in a row this proves that math is invalid.

Share with Sociable

Comments (229)

Trackback URL | Comments RSS Feed

  1. pete says:

    Ahem, you might want to fix the typo in title (and delete this).

  2. ploeg says:

    It’s as if they don’t understand that we’re dealing with probabilities (and not predictions that you can pull out of your ass).

    Or that both campaigns have an interest in portraying the race as close (to motivate their supporters to get to the polls).

    • thusbloggedanderson says:

      God does not play dice with the universe the election. His anointed must prevail!

    • Zifnab says:

      “Nate Silver said that there is only a 25% chance of two fair coins both landing heads up in a given set of tosses. Can we really trust his predictions if – after the Nov 6 coin toss – both coins show heads?”

      When Nate was on TDS, he talked about how he got his start with political modeling because of his love of internet poker. Then he joked about playing some of the pundits in a few games just to see how they would stack up. I suspect Silver will outlive these sad-sack pundits in the long run if only for the same reason that he got a job at the NYTimes to begin with. He’s playing the odds. They’re just betting where the wind takes them.

      Republicans loved Silver in ’10. Democrats love him in ’12. Silver will continue to be loved by the folks with the best odds of winning elections. These folks are, not coincidentally, the majority of voters.

      • NonyNony says:

        They’re just betting where the wind takes them doing their jobs as political operatives.

        Yeah, I’d believe they were “betting where the wind takes them” if their opinions didn’t so often match up with the political orthodoxy of the “Village”.

        They’re courtiers in Louis’s court and Silver is an outsider who doesn’t kiss the king’s ring in the right way. That’s pretty much all there is to it.

      • Stag Party Palin says:

        This. People who own casinos know what Silver’s talking about. The rest are just suckers.

  3. howard says:

    You had the gall to send me to politico?

    And to a moron?

  4. thusbloggedanderson says:

    Speaking of the sports-statistician Silver, weren’t there some baseball games over the weekend?

    • avoidswork says:

      And the evil librul city of SF crushed the soul and spirit of the conservative** Detriot.

      Naturally, some d-bags in SF had to go start fires on Mission and Market Streets, throw bottles and tip a car over.

      ((**at least according to Matt Lewis of The Daily Caller))

    • Keaaukane says:

      I was waiting for a World Series thread. Did anyone pick a SF sweep? And if they can’t predict a simple thing like baseball, why should we respect their opinion on complex items such politics, value of third parties, merits of Jewel, or the joys of walrus sex?

      • thusbloggedanderson says:

        Or the joys of walrus sex with Jewel?

        (Probably not much fun for Jewel, definitely not for the walrus, but of some satisfaction for third parties.)

      • spencer says:

        I didn’t see anyone predict a San Francisco sweep. I think I saw one guy pick the Giants in five, but most people who got the team right thought the series would go longer.

        As did I. But of course, I also picked the wrong team. So maybe I could write about baseball for Politico.

      • Fighting Words says:

        I read Sports Illustrated and the majority of their baseball experts picked the Tigers to win in 5 or 6 games.

        Coincidenally, I think these were the same people who picked the Rangers to beat the Giants in 5 or 6 two years ago.

        Also, regarding the 2-3 Detroit Tigers fans who read this site, this really was a much better series than the outcome. If Detroit played any other team than the Giants or A’s, I would have been rooting for the Tigers.

        • Linnaeus says:

          Giants played masterfully, and outside of Game 1, the games were pretty close each time. Fister, Sanchez, and Scherzer all pitched well enough to keep the Tigers in the game. It’s just that Giants’ pitching was a little better. Fielding was excellent, too.

        • timb says:

          To be fair, most sports writers feel if you can beat the Yankees, you have already won the WS. Half of them don’t understand why Detroit had to play a series at all

        • NorthLeft12 says:

          Excuse me Mr. Rasmussen Fighting Words, but I would guess there are a lot more than two or three Tiger fans on here. We are justifiably lying low/crying in our Pabst beer/and calling for Jim Leyland’s head on a platter.

          I’ll congratulate the Giants for doing to the Tigers what the Tigers just finished doing to the Yankees. A masterful performance by the Giants.

  5. Izzy says:

    This is awesome. “A Romney win, which Silver says will happen one time in four, will invalidate quantitative analysis! Avoid looking behind the curtain from which we churn out ‘analysis’ which we know will never put our jobs at risk, no matter how incorrect.”

    • TT says:

      They hate Silver because he doesn’t care about “likability” and “authenticity”, or who would be a better babysitter or drinking buddy.

      If nothing else, in a better world maybe Silver’s work will force the corporate lecture circuit to ask itself, “Now why are we dropping hundreds of thousands of dollars on pundits and ‘insiders’ who don’t have the slightest damn clue about what’s going to happen?” Or at least force newsrooms to reevaluate the value of having correspondents covering campaigns as nothing more than (not so) highbrow gossip columnists.

    • Unsympathetic says:

      I truly find it fascinating that anyone actually gives Republicans the benefit of the doubt.

      They lie. All the time. About everything.

      Of course the only reason they do this is because they want to lure in the idiots who want to vote for “the winner.” Just like all power-hungry fascists, everything must be filtered through a political lens, so Silver must not be allowed to stand.

  6. virag says:

    Why does Silver get all the play and not Sam Wang? Wang’s analytical model makes Silver look much more like a pundit than an apolitical nerd. Wang also has the benefit of having been accurate in the past. Is it the math? Wang’s analysis should be all the talk on the internets and the 24-hour political cable news, but he’s almost invisible. It is not just the news folks who live off the horserace, even if it doesn’t exist.

    • Stan Gable says:

      It’s the whole NYT platform thing.

      • Erik Loomis says:

        It’s probably also Silver’s baseball background, making him well-known among a wider group of people than Wang.

        • virag says:

          That does make sense. And Silver got a decent amount of television–well if MSNBC counts as television–exposure during the 2008 election season, but I am still a bit surprised that people with an outlet and an audience have not been talking up Dr. Wang’s work, especially on the Democratic side as his analysis has been good news for them for quite some time.

          • Richard says:

            Silver is a far better writer. His pieces, way before he got hired by the NY Times, were fun to read.

            • Richard Hershberger says:

              Although, alas, less so since he was hired by the Times. I hope he is well paid and wish him well, but he had a better blog when he was freelance.

              • (the other) Davis says:

                I wonder if that’s a result of NYT editorial control on the blog.

                • Richard Hershberger says:

                  The annoying convention of referring to a person as “Mr.” or “Ms.” after first mention is certainly NY Times house style. But more importantly, he used to have opinion pieces as well as analytics. He no longer does this. I suspect that this was a decision by the people that hired him.

          • arguingwithsignposts says:

            Silver used to post at the GOS. So there’s that. Sometimes it’s not what you know, but who. Okay, 99 percent of the time that’s it.

    • avoidswork says:

      Can you provide a link? Happy to add another poll to my bookmarks.

    • greylocks says:

      It could be because Wang currently has Obama at 91% probability and nobody believes him. I’m not saying they shouldn’t believe him, just that no one does.

      • virag says:

        It may seem unbelievable, but his methodology and analysis is solid, so a few interviews with some straightforward explanations could do wonders. Of course, that would make much of the nonstop horserace coverage look even sillier, so I guess that’s the real answer.

      • The Brain says:

        The basic reason Wang has a higher probability than Silver is that Silver gives more weight to interstate correlations. That is, if something happens today, it could affect multiple states in the same direction, not just one state. I suspect that many of the details of Silver’s model are extraneous complications, but on this key point, he has it better than Wang.

        • The Brain says:

          Pinky has been using my account again and now my cookies are a mess…. oops.

        • Green Caboose says:

          Wang addresses this, arguing that the complications are unnecessary. He also has good posts up now talking about how you can evaluate the performance of the different sites after the election is over (one key point: the evaluation should NOT just include the last prediction but the predictions over time) and another post defending Silver, even though he also cites his disagreements over methodology with Silver.

          On the 91% question, I think Wang’s correct, not because his methodology is necessarily better but because Silver’s Monte Carlo simulation methodology overstates the probability of unlikely events. For example, in 2008 he got all but one state right, true, but his probability of Obama winning was 95 or 96%. Sorry, but by the day before the election in 2008 Obama’s chance of winning was something like 99.9999%, not 19 in 20 or 24 in 25. But all those ultra-low probability scenarios he ran added up to an improbably high chance of a McCain win – and likely this is due at least in large part to his inter-state correlations, as each low probability scenario in a single state brought in other states and thus artificially increased McCain’s victory chances.

          Wang, on the other hand, has a much tighter probability distribution – it’s still a distribution with a reasonably wide range for the confidence interval, but not as wide or as flat as Silver’s.

          And given the polling data that makes sense. My concerns about an Obama loss are all based on vote suppression scenarios a la Florida 2000, when probably 40k or 50k more voters went to the polls intending to vote for Gore instead of Bush, or on some major flaw common to all polls (which is extremely unlikely). If anything, though, the major flaw is likely to be the lack of, or under-representation of, cell phones, and that will favor Obama.

          • mpowell says:

            There’s a pretty good argument for inter-state correlation, though. Months ahead of time, shifts in the national mood will show tons of inter-state correlation. And even the day before, all the pollsters use polling methodologies that makes some kind of assumptions about the demographics of voters that will show up and the average consensus could be off for an election. This would also lead to inter-state correlated shifts. I don’t know if Silver has the right level of correlation, but I don’t really understand how you could claim it doesn’t exist. Based on what I’ve read of Silver, I’d assume his model his based on historical data of correlated error between polling and election results.

          • ajay says:

            For example, in 2008 he got all but one state right, true, but his probability of Obama winning was 95 or 96%. Sorry, but by the day before the election in 2008 Obama’s chance of winning was something like 99.9999%, not 19 in 20 or 24 in 25.

            And you know this because…

        • Hob says:

          Most of Wang’s discussion of what he doesn’t like about Silver’s model is over my head, but the part about factoring in economic outlook made sense to me. If I understand correctly, Silver has built in a factor to say that if the economy sucks by X amount, voters will be Y% more likely to vote against the incumbent. But then he’s adding that to poll numbers, which presumably already take into account how the voters feel about the economy – in effect double-counting their discontent. That seems like a plausible reason for Silver to be underestimating Obama’s chances somewhat.

          • mpowell says:

            The economic stuff is removed as you get closer to the election. Which makes sense. It’s basically the argument that 6 months out you should consider how economic factors will impact voters minds when they think about things in October/November. And I’m sure he has a historical/evidence basis for including it, though I’m sure the exact weighting curve is artificial.

          • Discussion Question says:

            Silver factors in the economy earlier in the election period, on the theory that (1) ordinary voters will be influenced by *future* economic events, and (2) an index of half a dozen leading economic statistics can predict future economic events better than the voters.

            As the election progresses, the economy carries less and less weight in Silver’s model, and the weight drops to zero just before the election.

            So Nate agrees that how voters feel about the economy *today* is factored into the polls, but he thinks that economic statistics are a better guide for how voters will feel about the economy in the *future.* This seems plausible.

      • Leeds man says:

        The thing about probability is that it is not something to be “believed”. The credibility lies in the model, not the outcome. Of course, with general and (especially) pundit mathematical illiteracy, if Romney wins, both Wang and Silver will be discredited.

      • Davis says:

        Make that 97%. Sounds improbable, but like Schrodinger’s cat, once we look at it on Nov. 6, it will be either 100% or 0%.

        • Richard Hershberger says:

          I’m not sure we should find near certainty at this point improbable. If we take as a working assumption that nothing between now and the election is really going to change, then the problem is simply one of how complete is our information? We have a long history of polling, comparing it with actual results, and we have lots of current polling, at least in the places that matter. The idea that one week out we would pretty much know the result doesn’t strike me as all that improbable.

          • Green Caboose says:

            True. And I’ve seen some people criticize Silver based on his predictions of 2010 (he understated the number of house seats the GOP gained by roughly 10) or individual primary contests. The problem with those is that his model – or Wang’s or any additive model – gains strength when the polls are higher in numbers (as the sample size grows). For single seats there are far fewer polls so the model’s reliability is weaker.

            The problem is how to deal with noise and outliers. Rasmussen tends to get overrepresented in many polling aggregators because they produce so many of them. Some aggregators count all polls, others weigh more recent polls more heavily, and in both cases a pollster running highly frequent cheap robo polls is represented more often. Daily trackers can also be overrepresented for the same reason. On the other hand, daily trackers have the advantage of smoothing out the possibility of an outlier sample, whereas polls published every few weeks or so can have a really big miss from time to time – as we see sometimes from Pew. So a model that incorporates all polls equally and factors in historical data from that pollster, not the the latest, is likely to be more accurate than one that heavily weighs the latest – it will also be less noisy (as we often see from say the TPM averages).

            The other issue is general outliers and this is where using median instead of average works better IF you have a sufficient number of polls. The problem there is for states with less polling, where media can still be wrong – this is why there is some justification for Silver’s interstate correlations to smooth noisy bumps in one state.

            Finally, you have the issue with quality and reliability of pollsters and whether you take that into account. Wang basically dumps Gallup and Rasmussen and lets the others fall where they may. Silver uses a method that he’s hinted at but not specified for rating pollsters based on historical accuracy and their methodology (cell phones? robo polls? response rate? sample size?) so that seems more objective.

            If you have a rational model and enough data to work with you should be able to predict the outcome with high certainty, barring a major last minute polling swing as apparently happened in 1993 in Britain.

          • thebewilderness says:

            Except things are not the same. Suddenly the ability to vote needs to be factored in.

    • thusbloggedanderson says:

      Wang is the one who was going on about how the Dems were going to take the House this year, right?

    • Warren Terra says:

      There’s a pithy term for it that escapes me, but a huge part of it is just that Silver got there first, occupied the territory, and hasn’t underperformed in such a way as to undermine the advantages these gave him. Silver built his brand predicting the 2008 primaries, and had a strong following before the New York Times hired him. I don’t know when Wang started blogging his analyses, but I suspect he wasn’t doing so, or wasn’t reaching out so effectively, in 2008.

      • Stan Gable says:

        First mover advantage?

      • Kalil says:

        Not sure this actually works – Electoral-vote.com was the first highly popular ‘poll aggregator’, which FiveThirtyEight is essentially a highly refined variation of.

        • Warren Terra says:

          I could be wrong, but I think electoral-vote simply averaged available polls. Rightly or wrongly, Silver went well beyond that, while trying to explain what he was doing and why. And he offered frequent write-ups discussing his thoughts and concerns, and a whole bunch of snazzy tools.

          • thusbloggedanderson says:

            You’re right. Indeed, E-V relies mainly on the *latest* polls, not on the average.

            That’s the main reason I look at it, to get a different sense of the game from what other sites aredoing.

      • JL says:

        Yeah, I started reading Silver in early 2008 (and was delighted to see someone doing something more sophisticated than just averaging polls), and got a lot of my friends hooked on him then. I didn’t hear of Wang until this year. Not sure if this is a marketing issue or what.

  7. Chatham says:

    “So should Mitt Romney win on Nov. 6, it’s difficult to see how people can continue to put faith in the predictions of someone who has never given that candidate anything higher than a 41 percent chance of winning (way back on June 2) and — one week from the election — gives him a one-in-four chance…”

    I wonder if this genius thinks that difficult to see how anyone could get hurt playing Russian roulette.

    • Izzy says:

      I wonder if this genius thinks that difficult to see how anyone could get hurt playing Russian roulette.

      Sweet Jebus, this.

      Any passingly numerate person knows there actually is a test for Silver, and it ain’t “Romney wins.” If Silver’s prediction about the candidates performance in individual states relative to their national averages is off, that would be a massive blow to his model.

      If we wake up next Wednesday and Romney won the national vote by 1% and OH by 3%, that would be a blow to Silver. If Romney wins by %3 nationally and 1% in OH, not so much. It’s really not that complicated.

      • Steve H says:

        Silver comes up with percentages for each state. If he is wrong on a bunch of them, that’s a good sign his model is way off.

        There’s really no practical way to test his national model, since these elections only come once every four years.

        • Izzy says:

          Right, but that’s what I’m getting at. He’s making a mass of testable predictions about individual state results. It won’t be hard to compare his successes to other predictions. And yeah, the tiny sample size confounds. My point is just that his model is falsifiable, just not by “the candidate you said had a 25% chance one–how often could that happen, huh? HUH?”

    • Karate Bearfighter says:

      I think David Brooks would tell you that if someone gets hurt playing Russian roulette, you must have been wrong about how many bullets were in the gun.

    • Wido Incognitus says:

      LOL
      ALthough, I should point out the obvious that if your models are consistently unable to make accurate predictions, then it is foolish to defend yourself by appealing to probability.

    • John says:

      What’s remarkable about this is that nobody ever says anything about RCP’s unreliability, despite it’s genuinely awful track record.

  8. Bloix says:

    I have trouble believing Silver because I don’t see how he can model the effects of voter suppression.

    • Incontinentia Buttocks says:

      Why is voter suppression impossible to model? Voter suppression is not the result of random acts of political thuggery, but of laws that have predictable effects.

    • chris says:

      In theory, you could just make it part of your likely voter model. Not just “how likely is this person to attempt to vote?” but how likely are they to attempt AND SUCCEED.

      In practice, though, there’s very little data on the actual effect of voter suppression so you’d basically just be hand-hacking the numbers, which Nate is understandably reluctant to do.

  9. Silver or Wang– either one renders nitwits like Brooks irrelevant. As I listened to Cokie Roberts blither on this morning it occurred to me that there are quite a number of mainstream media outlets these days with no idea how clueless they are. Someone should tell NPR; at least the NYTimes has the wit to put Silver on the payroll.

    • Stan Gable says:

      NPR had Silver and Wang on for a joint segment last week. Granted, not the same thing as a recurring spot.

    • Green Caboose says:

      I can’t believe Cokie still has a job. In 1992 Gallup had a last minute bump to Bush then swung right back to Clinton in the week before the election. A literate person would have interpreted that as a statistical error. She decided that a bunch of people changed their minds out of fear of a liberal winning, then decided they couldn’t live with 4 more years of Bush so went back.

  10. Marc says:

    The right wing has really amped up the hate of Silver – just check the increasingly nasty comment page on 538 (or the personal attacks on him as corrupt or not “manly” enough or whatever.)

    It could be that they think that pretending to have an edge can make a win more likely, or they could be living in a fantasy world where they reject all evidence that they don’t believe. Whatever the cause, it’s yet another reason why we need the Republicans to lose this election decisively. It’s deadly to have such an unhinged party so close to controlling all branches of the government.

  11. whatever says:

    I disagree with your disdain for political pundits, especially David Brooks. Their jonb is to provide explanations that are based on observations and data, some of which may not fit into a model for mathematical analysis, and I think they do a good enough job at it. In comparison to Nate Silver, they have a role explaining why there are changes in the probabilities for certain results in elections.

  12. Cody says:

    “If you tell me you think you can quantify an event that is about to happen that you don`t expect, like the 47 percent comment or a debate performance, I think you think you are a wizard. That`s not possible,” Times columnist David Brooks, a moderate conservative, said on PBS earlier this month. “The pollsters tell us what`s happening now. When they start projecting, they`re getting into silly land.”

    Math, wtf is it?

    Also, using his own logic he has predicted a winner for the election. Thus he must be a fortune teller, because one cannot make predictions without 100% certainty!

    • Incontinentia Buttocks says:

      I wonder if Brooks thinks that meteorologists can only tell us what’s happening now and that, e.g., predictions of Hurricane Sandy’s track are “silly land.”

    • Anon21 says:

      He’s right that you can’t predict surprises, but Silver has never claimed to. He’s saying “given these polls, if nothing happens to shake up the race, here’s what’s likely to happen.” Given the qualifications, it’s not meant to be an absolute prediction of how the election will play out in reality. But then, pundits are generally no better at predicting specific gaffes, poor debate performances, or even which candidate on balance is likely to be helped or harmed by events like that.

      • spencer says:

        Given the qualifications, it’s not meant to be an absolute prediction of how the election will play out in reality.

        Exactly. But people like Brooks and most other pundits simply *cannot* deal with that kind of uncertainty. They need a definitive prediction and they need it right now. No nuance, no ambiguity. Their brains aren’t wired for those things.

        • greylocks says:

          I don’t think it’s that at all. I think it’s that they don’t understand statistics and probability, and it’s in their nature to reject what they don’t understand, rather to make an effort to understand it.

          They obviously don’t understand the difference between prediction and probability, or that poll results are not the same as the probability of winning.

          • Hob says:

            Very likely true, but irrelevant to the bit Cody quoted above. Brooks is saying that Silver is being silly because his model can’t account for unforeseen events. Silver, as Anon21 points out, has made it very clear that his predictions don’t account for unforeseen events, because how could you? Any prediction about anything has an implied footnote of “barring something unforeseen.”

            Brooks’s problem (if it is an honest problem and not just feigned ignorance) goes way deeper than not understanding math; he’s denying that any statement about what’s likely to happen, quantitative or not, can ever be meaningful. By the standard he’s using here, I shouldn’t say that the Giants are taking home the trophy even after they won the last game, because maybe they’ll all spontaneously combust.

            • Pseudonym says:

              But isn’t part of his probability model considering the likelihood of unexpected events swinging the electorate? Isn’t that part of the difference between the forecast and the nowcast?

    • Tide goes in, tide goes out. Sun, moon. Stars.

      You can’t explain that.

  13. Walt says:

    The comments are surprisingly anti-pundit, and pro-Silver (or more accurately, pro-statistics.)

  14. Jerry Vinokurov says:

    The whole Silver thing reminds me of a good joke I heard recently (which I’m sure many of you know already): when you tell Republicans that the polls disagree with them, they want to kill the pollster; when you tell Democrats, they want to kill themselves. If Nate Silver were telling me that Romney had a 75% chance of winning the election, I would be looking for the best way to drink myself to death afterwards, but I wouldn’t be going around saying, essentially, “math is a conservative lie.”

    • parrot says:

      great joke, and proof that despite my voting habits, i am in fact a closet rethug … how am i going to look at my wife and dawgs anymore … i feel so cheap, used …

    • Linnaeus says:

      Conservatives have embraced Lysenkoism. Their only real complaint about him is that he was a communist.

    • Green Caboose says:

      That “joke” was actually what a pollster supposedly said, anonymously, to a pundit this year. I’m sure if you google it the source will show up.

      Kill the messenger has long been the conservative philosophy. I remember back when I slummed on conservative sites and I saw this in spades. For example, you may recall that Fox News’ last presidential poll in 2004 favored Kerry. The conservative commentators were virtually unanimous that Fox was lying and that they were now as bad as the liberal media.

      By the way, this is why any attempts by organizations like CNN or Newsweek to eat into the Fox News audience is utterly destined to fail (if that is in fact what they are doing, and not just trying to present things from their owner’s p.o.v.). If you don’t tell the modern conservative what they want to hear ALL the time you are considered a liar and a liberal. Just look what they did to Frumm.

      • Yeah, that’s right. I should have said “anecdote,” rather than joke; it was relayed to me during the course of a lecture I attended, which was given by David Barker, a Pitt prof. who studies polls. For whatever it’s worth, his prediction for the race was something on the order of Obama being a 2:1 favorite, though this was also two weeks ago.

        • parrot says:

          well, this was the first i’d heard it … my cluetrain has no engine, only caboose … it had a mort sahl vibe of yesteryear … de gustibus non est disputandum

  15. Anon21 says:

    I don’t know why, but today, for the first time, I’m convinced Obama is going to lose. The steady march of those national polls just feels too much like 2004, when most signs pointed to Bush but I fooled myself into only seeing the stuff that was good for Kerry.

    And I do realize that Obama is in a stronger position now than Kerry was then. But the relative consensus in the national polls around a small Romney lead has me scared to death.

    • Jamie says:

      Correct me if I’m wrong, but isn’t the small Romney lead mostly a function of Gallup and Rasmussen having very different ideas about the makeup of the electorate than other pollsters?

      Now, granted, Gallup and Rasmussen could be right, but I don’t think it’s fair to say that the “consensus” is a Romney lead.

      • Anon21 says:

        The ABC/WaPo tracker has also been showing a small but consistent Romney lead (although today it’s a tie, which honestly does make me feel better). The last AP poll showed R+2. Last SUSA was Romney +3.

        All I know for sure is that the RCP national average has been stuck on Romney for over a week now. Maybe that’s an artifact of selective inclusion (although I can’t fault them for excluding, say, the online tracking poll) or maybe it’s something real, but it worries me.

        And at the same time, the news out of Ohio has been getting a little less encouraging. Those spectacular early voting numbers are being debunked as less reliable than the county-level reports on actual ballots cast, which are much more ambiguous. Today saw the first poll in over a week with Romney holding even a nominal lead in Ohio.

        I dunno. It’s entirely possible I’m overreacting to minor changes in the news environment, but I’m honestly freaked out. To some degree, I kept expecting the national polling to revert to a small Obama lead at some point after Denver and it just… hasn’t.

        • Green Caboose says:

          OTOH, Rasmussen has Romney’s lead down to 2. Rasmussen historically shifts away from the GOP near the actual election so their final poll is relatively close (see also: Zogby, whose reputation is now so bad he changed the firm name).

          The argument about cell phones, which was just emerging in 2008, is now significant. Over half of voting age Americans are either cell-phone only or cell-phone primary – and those people are missed in almost all polls being run today. And yes, their demographics are very pro-Obama … minorities are much more likely to be dependent on cell phones now (which, with pay-as-you-go plans, are cheaper than landlines and basically essential to modern society) as are young people. Also much more so in urban society versus rural where coverage is still spotty in many areas.

      • Too Close to Call says:

        Obama’s lead is collapsing in OH now too. First a tied poll, now a poll with Romney ahead.

        And it seems the map is expanding to include Pennsylvania and even Minnesota.

        • Anon21 says:

          No, I don’t even buy that Minnesota and Pennsylvania are in play in any meaningful sense. That’s a Republican confidence play. But until I see a few more Ohio polls showing Obama leads of +2 – +4, I’m not feeling too comfortable about the outcome.

        • thusbloggedanderson says:

          Guys & gals, the “tied” poll is an outlier. If Obama’s up two and the MOE tends to +/- 3, then you’re going to get some of those.

          As for “a poll with Romney ahead,” yes, if you call only landlines and tweak your LV screen, you can get a result with Romney ahead.

          • Too Close to Call says:

            The poll with “only landlines” has never had Romney up in Ohio before. In the last poll, done by the same pollster, Obama was in a much stronger position.

            So the momentum is in one direction–to Romney.

            And the early voting in Ohio is overblown.

            • thusbloggedanderson says:

              Says ONE poll? Whatever.

              The purpose of averaging the polls is to avoid freaking out due to one or two wacky ones.

              But from your other comments, I see that the facts are not your strong point.

              • Too Close to Call says:

                There have been two, very recent polls of OH from non-partisan firms.

                One in-state poll–considered to be the gold standard IIRC–showed the race all tied up.

                Rasmussen, one day later, showed Romney up by two.

                That’s called “momentum” and it’s on the Republican side.

                • thusbloggedanderson says:

                  Two data points = momentum?

                  Look up “fluctuation.”

                • Anon21 says:

                  It’s not really called momentum. And it’s not clear that there’s been a move to Romney at all; if the race is truly Obama +2, we would expect to see some polls showing Romney +1.

                  But… this close to the election, there’s less time to sort out what’s a genuine shift and what’s float in the margin of error. And that’s why the Romney +1 poll has me freaked out. I concede it’s not rational, it’s just where I’m at.

                • Warren Terra says:

                  Please parse this for me: Rasmussen is one of the two “non-partisan firms” you’re referring to?

                  OK, then.

                • Anon21 says:

                  Rasmussen is a non-partisan pollster. Don’t fall into the trap of dismissing their polls because they have a small Republican house effect. They’re genuinely trying to get it right, and their track record is decent, if not outstanding.

                • thusbloggedanderson says:

                  They’re genuinely trying to get it right

                  Which is they’re two points off in the GOP’s favor, on average?

                • Anon21 says:

                  Just like PPP has had a Democratic house effect in the range of two points.

                  These house effects usually arise from genuine differences of opinion about how to best model the electorate. Rasmussen’s opinions about that haven’t always been borne out, but by no means are they outside the bounds of reasonable, good-faith methodological disagreements, and by no means do they invalidate its polling results.

                • thusbloggedanderson says:

                  “Just like PPP has had a Democratic house effect in the range of two points.”

                  Right, and they too are robocallers, so I don’t take them very seriously either. Is that an argument *for* Rasmussen somehow?

                  Go to any state you’re curious about at HuffPo, look up their polls, and then figure out which ones are robocalls or internet. Throw those out and average the rest.

                • thusbloggedanderson says:

                  And while we’re on polls: ANYONE got a clue why TPM, alone of the various sites, thinks Michigan is a tossup with only +1.2 for Obama?

                  Even RCP thinks Obama’s up 4 there. HuffPo, +7.

                • Methodology here.

                  Q3: Is there any editorial input in the averages? Can they be adjusted in any way?

                  A3: No. Other than rejecting clearly flawed or fraudulent polls (see A1), all polls are treated equally, and there is no editorial input whatsoever.

                  The moderately screwy remains after the obviously screwy is tossed.

                • thusbloggedanderson says:

                  Yeah, but even looking at their own numbers, I can’t figure that average out.

                  RCP doesn’t discard any numbers either, and they aren’t getting 1.2.

                • Good call on your part, really…gotta go a ways back to get Romney in a winning position. I wonder how much “tracking” the polltracker does against history in this case.

                • There have been two, very recent polls of OH from non-partisan firms.

                  Uh, actually there have been seven very recent polls of OH (within the last week). The only one that doesn’t show an Obama lead is the very partisan Rasmussen poll.

                • They’re genuinely trying to get it right, and their track record is decent, if not outstanding.

                  Rasmussen’s final polling has a decent track record. Rasmussen’s polling during the election, on the other hand, has a record of being consistently skewed towards the Republicans, only to “discover” significant movement towards the Democrats in the final week or two that brings their numbers into line.

            • DocAmazing says:

              the early voting in Ohio is overblown consists of actual ballots.

              Fixeded.

              Remember when Republicans loved early voting results because they indicated the will of retirees and other mostly-Repbulican groups?

  16. Too Close to Call says:

    Reality check folks: Nate Silver was way, way off in 2010–20% off in fact.

    He got lucky one year.

    This race is a coinflip period.

    • Anon21 says:

      If you mean 2008, I don’t think he got lucky so much as that was an easy race to predict. People talk in reverential tones about how he called 49 of 50 states correctly, which is great, but they seem to forget that only in about 6 to 8 states was there any doubt about the outcome.

      • Too Close to Call says:

        He was way off. He said Republicans would get +53 seats. They got 63 seats. That’s not close.

        • Jerry Vinokurov says:

          Shorter Too Close: I do not understand confidence intervals.

        • Jerry Vinokurov says:

          And this of course doesn’t even take into consideration the fact that data on house races is more sparse than data on the presidential race.

          • Too Close to Call says:

            He was wrong about the number of seats. Period. His precious model was way off.

            • Jerry Vinokurov says:

              Are you seriously that fucking dense or what? If I give you sparse data, no matter how good your model is, it is not likely to give you the right answer. Yes, he got the number of seats wrong, but the only people who put a period at the end of that sentence are people who don’t understand basic statistical concepts, like confidence intervals.

              Or, allow me to summarize your objections more succinctly: hurf durf.

              • arguingwithsignposts says:

                Don’t interrupt him when he’s fucking that walrus. He might rage all over you.

                • Jerry Vinokurov says:

                  Your handle is very appropriate in the context of this debate…

                  I guess Republicans already think that physics and biology are a liberal plot, so why not math too?

              • Too Close to Call says:

                I see Obama’s slip in the polls is starting to get to some of you and making you quite defensive.

                Don’t worry, it’s still a possibility that Obama pulls off a very narrow win. But looking at the fundamentals of this race, especially at where the momentum is, and it looks like Romney’s to lose.

                • JL says:

                  No, arguing with people who don’t understand math is starting to get to some of us, especially those who work with math for a living (I don’t know if that applies to Jerry, but it applies to me).

                  Seriously, though, as a Dem, I’d prefer that Dems think the race is close, so that they actually turn out. Why would I (or Silver, a liberal) want to lure Dems into a false sense of complacency?

                • (the other) Davis says:

                  Don’t worry, it’s still a possibility that Obama pulls off a very narrow win. But looking at the fundamentals of this race, especially at where the momentum is, and it looks like Romney’s to lose.

                  Who needs “statisticians,” with their “data” and “models,” when you can get your predictions from some random blog commenter? Do you do stock picks, too? If so, count me in for your newsletter.

                • thusbloggedanderson says:

                  How’s that “your disagreement with me indicates you fear I am correct” trick working out for you around the internet, TCTC?

                • Jerry Vinokurov says:

                  I find myself in these types of arguments all too frequently, alas. It’s not even particularly hard math; just basic stats and probability.

                • I see Obama’s slip in the polls is starting to get to some of you and making you quite defensive.

                  Personally, I’m trembling like the load-bearing walls of the Saginaw Plaza Hotel.

            • JoshA says:

              So in 2008, he got 48.66 of the states right (he got Indiana and Nebraska’s 2nd CD wrong, predicting they would be close but go McCain when they went Obama).

              He then proceeded to correctly call that Scott Brown would win the special election for the MA Senate seat.

              In 2010, he correctly predicted 34 of the 37 Senate races. The ones he got wrong were Colorado and Nevada (both times picking the R challenger when the D incumbent won) and Alaska (picking GOP/TP nominee Miller over incumbent Republican turned Independent Murkowski).

              He didn’t do a seat-by-seat prediction in the House, and gave the GOP a predicted pickup of 23-81 with a average of 54.

              In short, he has a remarkable record of accuracy, and each of the times he analyzed an individual race and was wrong he incorrectly favored the more rightwing candidate.

            • (the other) Davis says:

              He was wrong about the number of seats. Period. His precious model was way off.

              You’re just as clueless about probabilities as the dude who wrote the Politico article, I’m sad to say.

        • Marc says:

          Those are mighty thin reeds that you’re grasping.

        • John says:

          That’s 10/435, not 10/63, you idiot.

          • Cheap Wino says:

            This makes me laugh. He comes around blathering as if he has a clue and made this slap-in-the-face obvious error.

            Correction, he probably didn’t make the error, the dolt that wrote what he’s parroting did.

    • thusbloggedanderson says:

      Yah, he *overestimated* the GOP success.

    • greylocks says:

      Old poker player’s joke applies here:

      “I figure the odds of making my flush are 50-50. Eithet I make it or I don’t.”

  17. Too Close to Call says:

    Anybody that looks at this race with an objective lens can see that Obama is slowly but surely losing ground this week.

    Nobody has ever lost the Presidency when they’re over 50% in the Gallup tracker. Ever since they started polling. Romney’s at 51%. Think about that.

  18. Too Close to Call says:

    How can liberals dismiss Rasmussen and then turn around and make fun of conservatives who talked about “unskewed polls”? Looks like both sides do it when it benefits them.

    • Murc says:

      Because… the conservatives who are talking about unskewed polls are in fact talking about, really skewed polls?

      I’m pretty comfortable making fun of them.

    • Casey says:

      Because we liberals actually know how math and statistics work?

      Silver’s predictions are based on the statistical principles that, you know, led to putting a man on the fucking moon, inventing the internet, etc.

      The “unskewed” polls approach is based on assuming those same statistical principles are complete horseshit when they show a Republican losing and 100% unquestionable when they show a Republican winning.

      You know, science, reason, logic. All that stuff with a well-known anti-Republican bias.

      Anyway, statistics is neat! What level of mathematical education do you have, exactly? If you can’t understand stats (for instance demonstrating above that you clearly have no fucking clue what a confidence interval is), maybe leave it to the people who do understand it?

    • How can liberals dismiss Rasmussen and then turn around and make fun of conservatives who talked about “unskewed polls”?

      For the same reason that I can look at Kenin Garnett and call him tall, while making fun of someone who looks at Kevin Garnett walking down the street and declares everyone around him to be short.

  19. IM says:

    bookmark this, libs!

  20. Joseph Slater says:

    As of this writing, InTrade has Obama as about a 62-38 favorite, and Real Clear Politics — hardly a liberal site — has, in their “electoral college without leaners” calculation, Obama at 290 electoral votes. So, while Obama isn’t a sure thing, much better to be in his position now than in Romney’s.

    And if you want non-data-driven analysis, it’s worth pointing out that Romney has to deal with the fact that he called FEMA and federal disaster relief “immoral” right when the storm of a lifetime is battering much of the east coast. All in all, good news for John McCain.

  21. Steve S. says:

    it’s difficult to see how people can continue to put faith in the predictions of someone who has never given that candidate anything higher than a 41 percent chance of winning…For all the confidence Silver puts in his predictions, he often gives the impression of hedging.

    Then Brooks: “The pollsters tell us what`s happening now. When they start projecting, they`re getting into silly land.”

    Sweet Jesus. If I may be allowed a probabilistic statement, there is a 74% chance that these two went into journalism rather than the sciences because they are idiots.

  22. wengler says:

    If pundits believed in probabilistic outcomes, then global warming would be freaking them the fuck out.

    Luckily they don’t and are comforted by their blanket of ignorance.

  23. Anonymous says:

    Nate has a background in modern baseball statistics, a/k/a sabermetrics. Back in the 1980s sabermetricians like Bill James were as despised by the sportswriting fraternity as Nate Silver is loathed by our Beltway punditocracy.

  24. Karen says:

    So, is the President going to win? I’m suddenly petrified.

    • Vance Maverick says:

      I don’t think anyone’s model accounts for voters turning to stone. Please don’t, and please discourage your friends as well. Even with that, the best predictions are only probabilities.

      • Karen says:

        I live in Texas, so my vote, which I’ve already cast for the President, was a complete waste.

        • Bob Loblaw says:

          If you actually believed that, you wouldn’t have wasted your time. You may think you believe it, but deep down I don’t think you do.

          Think about it, Karen. Why should a single additional vote for the loser in an election be any more of a waste than a single additional vote for the winner?

  25. The wingnuts are always saying this election is looking like 1980. Or their fantasy of what happened in 1980, anyways.

    Turns out, Carter never led after the end of May.

    Reagan was not a come from behind win. Although the race tightened at the end, it was never really in doubt.

    This year, Romney has never really led in the polls. If there’s any comparison to 1980 (which I doubt) it’s not favorable to Willard.

  26. Reilly says:

    I almost stopped sipping my latte when I read this:

    For this reason and others — and this may shock the coffee-drinking NPR types of Seattle, San Francisco and Madison, Wis…

    Isn’t there some sort of Pundit Manual of Style to keep amateurs like Byers from placing a general, innocuous behavior where the specific pejorative characterization should be?

  27. The anti-Silver punditry is so weighted to the right side of the political spectrum that it makes me think that something else is going on.

    The Romney campaign wants to create the image that he has momentum – you might have heard that word “momentum.” It means that Mitt Romney just. wins. football games. It’s all he does! – and Nate Silver’s numbers-based analysis throws cold water on that story.

    I agree that pundits love a close-race narrative, but this looks more like Republican tactical b.s. more than a gatekeeper reaction.

  28. cpinva says:

    the presidency was lost for the republicans, the moment the “klown kar masquerading as a primary” began. silver simply quantified it into an easily digestible bit of cooked potato.

    “Pundits Are Aghast” should really be the name of an alternative rock group, but it will suffice as a description of the reaction by the “Usual Suspects” in political pundit land, to mr. silver’s pretty accurate assessment of their boy’s chances on nov. 6. the horror, the horror!

    what really pisses them off, is that, if silver proves as accurate as he did in 2008 & 2010, they’ll be permanently out of work on nov. 7th. who’ll need them, when all they have to do is check with silver? lost ratings, lost ad revenues, lost jobs. money, money, money, moneyyyyyyyyyy!

  29. SpaceSquid says:

    The fact that David Brooks thinks accurate statistical inference is akin to wizardry is a surprise to very few people, and the fact that he’s being a dick about it a surprise to fewer still.

Leave a Reply

You must be logged in to post a comment.