Wednesday Highlights

Good morning. I hope all slept long and well. I didn’t … someone put the thermostat on “hold” and the house stayed warm, which under the covers meant hot … and it never cooled off to comfortable, i.e., under 60.

  1. Baby keeps self safe.
  2. My two election day links for the rest of the week … and here’s the other.
  3. On wealth, words from the 5th century.
  4. Curiosity.
  5. Solar UV.
  6. Not the standard POW fare.
  7. Wow. Just wow. …
  8. this is not unrelated.
  9. I think our discussions on these haven’t stopped. Hopefully. At least our prior conversation established that using a drone is an act of war on the country in which it is used.

43 Responses to Wednesday Highlights

  1. So, using polling data, Nate Silver correctly predicted 50 out of 50 states this year and 49 out of 50 states this year. Questions.

    1) How do you explain this if polls are just cricket races?

    2) What do you think of the credibility of the Republican pundits and bloggers (including the one you linked to a couple of days ago) who were bashing Nate Silver as biased and saying that Romney would win, possibly by a lot?

    3) In short, is reality clashing with your prior beliefs going to change your beliefs at all?

  2. (49 out of 50 4 years ago, obv.)

  3. http://www.slate.com/articles/news_and_politics/politics/2012/11/pundit_scorecard_checking_pundits_predictions_against_the_actual_results.html has a nice little dart board chart showing how close Dem and Rep pundits were to predicting the outcome. It does seem, as a pattern, that Republican orientated pundits are off by a larger margin than Democratic ones.

    I think this works with my thinking the right is more geared toward loyality think whereas the left is more geared towards geek think. Lefties want to geek out on the polls and get it right. Notice all of them made it on the board, except for Jim Cramer with an absurd prediction of 440 votes for obama. Right wing pundits seem much more likely to be off the board, let alone far from the bull’s eye.

  4. Thanks for the link, Boonton. That is great.

    I agree with your loyalty vs. geeking out distinction, although both phrases aren’t quite on the nose. Rather than “loyalty” it’s “believing whatever supports your guy” and rather than “geeking out” it’s “trying to figure out the truth regardless of who it supports.”

    The distinction was on full display during the debates. After the one that Obama lost, everybody on the left — I mean EVERYBODY on the left, as far as I can tell — acknowledged that as the ground truth and talked about why it happened and how Obama could come back, etc. The one that Romney lost bad, it seemed like a huge number of people on the right couldn’t admit it, even to themselves.

    It’s nice to see the right’s bubble hurting only themselves for a change. I flipped over to Fox for a few minutes last night for the shadenfreude and saw that the on-air talent was arguing with even FOX news’s experts about whether Ohio should have been called. The denial is strong with them.

  5. An interesting observation (that ties into all this) that Nate Silver made last night was that in the Romney states, Romney tended to win big. This might be why national polls were showing such a toss up while state by state electorial map based polls were more favorable towards Obama. Winning Texas @57% is nice but Romney would have been better off winning it with 52% and gaining that 1.5 M votes or so in states like PA, FL, and Ohio.

    File this under the ‘inside the bubble’ bias. I’ve written before this is not always a bad thing. No football team goes out on the field with a ‘realistic assessment’ of their chances. They take the field hyped up on a speech that they are the greatest team ever. Those who geek out on the stats may make great bookies but won’t make a great coach or cheerleader.

    True Obama’s margins in NY were larger (63%) as well as CA (59%) but the logic holds. At this point the Democratic party seems to own NY and CA when it comes to Presidential races and the GOP owns Texas. The battle is over the moderately sized states inbetween those big ones (Ohio, FL, VA, NC, PA). There the GOP needs to capture voters who don’t make it a point to listen to only Rush Limbaugh and get their ‘liberal POV” from Fox News. To do that they will need fewer cheerleaders and more analysts.

  6. JA,
    As I noted, I haven’t compared state/state polling data by the various polls to results. I’m assuming that a lot of states/polls have results outside their margin of error. I’m unclear how the Mr Silver example helps you point of view, perhaps you could explain.

    If I were to tell you, science has determined the mass of the electron to be .511 Mev this is quite different than your polling results. This would be equivalent to saying “the mass of the electron is reliable when Mr Silver measures it. Ignore the masses you get from Gallup &c.” This is a demonstration that this is not science. It is art. Some guys can pick winners in cricket races. The “know” the crickets. It is however not something they can write down and describe to, say, a computer how to pick a winning cricket, nor can they easily explain to another guy what he is doing wrong.

    How then is reality clashing with my beliefs. Please explain where you think the problem lies?

    Just curious as I have never seen Mr Silver’s projections, when you say he is right 49 out of 50 … do you mean to pick the winner or pick the margin and stated the error of his estimate?

  7. Boonton,
    Test your loyalty hypothesis. Look at pundit predictions in the last Pres. election that the Dems lost, i.e., Bush vs Kerry. Did the Dems predict loss? Or did was their prediction mapping more akin to this years?

  8. I’m unclear how the Mr Silver example helps you point of view, perhaps you could explain.

    He is not a pollster. He is a statistician who combines polls from all over the country and weights them according to various factors. If you were right about polls being just cricket races, then he would have to be extraordinarily, unbelievably lucky to have compiled the record of success he has.

    I knew going into this conversation that you were never going to admit you were wrong, but it’s of course still frustrating to see. Polls work. They have problems, but they are far more than cricket races, pundits, and bloggers. It is sad that you cannot either see this or admit that you were wrong.

  9. As I noted, I haven’t compared state/state polling data by the various polls to results. I’m assuming that a lot of states/polls have results outside their margin of error. I’m unclear how the Mr Silver example helps you point of view, perhaps you could explain.

    I’m unclear myself exactly how Mr Silver does his polls but I believe he is essentially averaging polls together. If poll errors are essentially random then averaging them together should give you an excellent estimate, which it seems like he achieved. Now he also seems to have done some scenaro calculations to find ‘how many ways’ Obama and Romney could have won to calculate probabilities not only for each state but the election as a whole.

    If I were to tell you, science has determined the mass of the electron to be .511 Mev this is quite different than your polling results. This would be equivalent to saying “the mass of the electron is reliable when Mr Silver measures it. Ignore the masses you get from Gallup &c.” This is a demonstration that this is not science. It is art. Some guys can pick winners in cricket races.

    Isn’t it more like when we put together multiple lab results we find their average of averages is 0.511 Mev?

    Test your loyalty hypothesis. Look at pundit predictions in the last Pres. election that the Dems lost, i.e., Bush vs Kerry. Did the Dems predict loss? Or did was their prediction mapping more akin to this years?

    Good point, not sure how easy it will be to find that.

  10. An interesting fact looking at the dart board, all of the pundits predicted ‘their guy’ would win. The one exception was Ross Douhat who predicted a very slight Obama win (271). Curiously he is the right wing pundit who writes for the NYT. Perhaps having to work around liberals or write for a more liberal audience caused his bubble to be pricked producing a more accurate prediction rather than a ‘feel good’ one. Jim Cramer likewise was way, way off but I’m not sure he’s a ‘liberal pundit’ nor do I think CNBC is that liberal. Interestingly MSNBC pundits weren’t too far off even though MSNBC is thought of as a left wing version of Fox.

  11. Boonton,
    One description of what he does is a Monte Carlo simulation, running “test elections” over and over millions of times. I don’t know how he builds his models, but that isn’t exactly as simple as “averaging” polls together. As noted earlier, to proper average different polls you have to get data from the poll that isn’t normally available in order to intelligently average them together.

    Isn’t it more like when we put together multiple lab results we find their average of averages is 0.511 Mev?

    No single measurement should be such that the actual value falls outside of their error bounds. If it does, that either indicates a methodological error or that their error estimates are flawed … or there is new Physics involved.

    JA,

    I knew going into this conversation that you were never going to admit you were wrong, but it’s of course still frustrating to see. Polls work. They have problems, but they are far more than cricket races, pundits, and bloggers. It is sad that you cannot either see this or admit that you were wrong.

    My only actual prediction was that well respected polls would be found in which the actual result fallws outside of their error estimates. Apparently you think I made a stronger claim.

    If you were right about polls being just cricket races, then he would have to be extraordinarily, unbelievably lucky to have compiled the record of success he has.

    You didn’t answer my question. Did he pick winners in states or results + error?

    If the former, apparently you have a low bar for “unbelieveable”, e.g., picking Kentucky or Texas for Romney isn’t unbelievable luck. All the bloggers, pundits, pollsters and likely Mr Silver estimated only a few states the outcome was likely to be close.

  12. Boonton
    Consider these remarks separating pollsters and pundits.

  13. My only actual prediction was that well respected polls would be found in which the actual result fallws outside of their error estimates. Apparently you think I made a stronger claim.

    I’m talking about your characterization of polls as “cricket races.” You’re attempting to shift the goalposts with all this nitpicking about margins of error and Nate Silver’s precise methods. The point is that polls provide valuable information and despite any flaws are pretty accurate — more accurate than pretty much anything else.

  14. Ok I just saw this on Nate Silver and the margins of error:

    Nate Silver did well. Sure, he got all 50 right, but he predicted vote proportions for Obama and Romney in every state, and for forecasters like me, the better measure is how closely he did on the specific margins rather than right or wrong. You could forecast DC as 51% Obama and be “right” on the electoral college, but that’s a terrible prediction (it was actually 91.4%).

    So here’s how well Nate Silver did:
    – Got the winner in all 50 states and DC
    – Got both Obama and Romney vote shares within margin of error for 98 of 102 predictions (Hawaii and West Virginia) — typical margins of error assume 19 times out of 20, or just over 95 of 102.
    – 1.5% error on average across all states (many of which had little polling) — his suggested margin of error was 3.7%
    – 0.77% error on average across 10 battleground states — his average suggested margin of error was 2.9%

    That’s pretty fuckin’ good. His biggest inaccuracy was understating how accurate his predictions were.

    Here’s an album of charts of how he did, with his margins of error included. I haven’t seen it anywhere, so I had to make it.

    From http://www.metafilter.com/121588/we-choose-to-have-an-election-thread-and-and-do-the-other-things-not-because-they-are-easy-but-because-they-are-hard#4665839

  15. JA,
    And how many times do I have to recount the drunk under the light-post searching for keys joke/analogy? They may be “more accurate” than anything else, except that using the term “accurate” is a mistake.

    It occurs to me from what I’ve heard is that Mr Silver didn’t actually make any actual predictions. He offered numbers like “Obama has a 60% chance of winning state X”, which means if Mr Obama did not win the state he’d still be technically right, just as the weatherman is putatively “correct” when he says there’s a 60% chance of rain and it doesn’t rain. Run the “experiment” 10k times and you may discover that he was right. Except you can’t run either the election or the weather for the day 10k times so the probabilistic interpretation of the result is useless.

    I asked you before and you didn’t reply, so am I right in the assumption that Mr Silver offered no error bars? Did he give spread predictions or just pick winners?

    Didn’t you learn in school that an experimental number without error bars is meaningless?

  16. Did you miss my previous comment? It includes charts with error bars.

  17. JA,
    It was in moderation and I hadn’t seen it.

    Did you see this? How could that have ever happened? Imagine that, races that had “surprises”. Gosh surpises in elections should be unheard of, if this science of polling exists.

    I don’t know what this sentence means: “typical margins of error assume 19 times out of 20, or just over 95 of 102.” Is that just an awkward way of saying 5% error bars?

  18. So do you maintain your stance that polls are just cricket races?

  19. JA,
    How to you reconcile not-cricket-races with polling “surprises” and many other polls which have results well outside their margin of error?

  20. JA,
    I might also chide you for accusing me at the same time of dismissing the primary relevance of polls and at the same time not dismissing poll interpretations given by bloggers like Borepatch? Do you really think I was saying at the same time that one should regard as rock solid one particular set of poll results and at the same time say polling is an inexact inconsistent unreliable methodology?

  21. I don’t really want to wrestle with you about all the details. Again, if you are genuinely interested, you can use google. I just want to point out that the people those of us on the left trusted were using math and statistics and reality while those you on the right trusted were using wishful thinking and gut feelings.

    As usual.

  22. JA,
    Except that in past elections where the Dems lost … those on the left were the ones trusting to wishful thinking and gut feelings and … what? the right was using math and statistics. Boonton pointed out “one” pundit predicted the other side to win. Likely this is quite similar in past elections, e.g., Kerry/Bush … that Dem pundits were not predicting a Kerry loss. I also linked a post in this comment thread which discusses the difference between pollsters and pundits.

    Apparently Mr Silver is the only pollster in existence. We’re not wrestling details. You claim polling is accurate. I point out many times in which polls are “surprised” and results fall outside of error bounds. When this happens in science, red flags go up. Yet, here it does not. Why is that?

    Tell me this. In the next election cycle in two years, will the published and now well estbalished methods used by Mr Silver be used universally by all polling agencies? If not. Why not.

  23. One description of what he does is a Monte Carlo simulation, running “test elections” over and over millions of times. I don’t know how he builds his models, but that isn’t exactly as simple as “averaging” polls together.

    Well suppose your average of polls says Ohio and Florida are 51-49% in favor of Obama with a 3 point variation. In that case it’s possible for Obama to win one, two or none. Running millions of test elections, you could scoop out how many times one would expect to see combinations like Romney winning one of those states (which gives him a shot to still win) versus Obama winning both or one. From there you produce millions of elections. In maybe 99.9% of those elections Texas always goes to Romney and CA always goes to Obama but the breaks in the middle states tell you how often the chips will fall in Romney’s favor versus Obamas. Hence his probability of an Obama win at 90% long before the election actually happened.

    I’m talking about your characterization of polls as “cricket races.”

    Maybe it’s me but what exactly is this analogy? What are ‘cricket races’ and what would it mean if polls were like them?

    It occurs to me from what I’ve heard is that Mr Silver didn’t actually make any actual predictions. He offered numbers like “Obama has a 60% chance of winning state X”,

    As cited previously, he made other predictions. But just to focus on this one aspect, if in ten states you say that there’s a 60% chance person X will win then one would expect your call to be right about 6 times or so. It’s not necessary to run thousands of elections in multiple universes. A single election can be broken down into multiple sub-elections where you can test your methods multiple times.

    I don’t know what this sentence means: “typical margins of error assume 19 times out of 20, or just over 95 of 102.” Is that just an awkward way of saying 5% error bars?

    I believe so.

    How to you reconcile not-cricket-races with polling “surprises” and many other polls

    Gambling is built on a very rigerous mathetmical principles of probabilities. It would be very surprising then if a well run Las Vegas casino was bankrupted by someone who wins 50 times in a row constantly betting all his winnings again and again. But at the same time the mathematics of probability say it can happen and if you had enough casinos, enough gamblers or enough time it would be surprising if it didn’t happen sooner or later!

    Except that in past elections where the Dems lost … those on the left were the ones trusting to wishful thinking and gut feelings and … what? the right was using math and statistics. Boonton pointed out “one” pundit predicted the other side to win. Likely this is quite similar in past elections, e.g., Kerry/Bush … that Dem pundits were not predicting a Kerry loss.

    I think Silver’s success demonstrates that to really do polling right you need a lot of resources and a lot of intelligence. You need multiple independent polls and you need to combine their results carefully. If you just want to look at one or two polls it’s very easy to fool yourself into thinking your side will win. Building a set of ‘all possible outcomes’ it’s possible to see paths to your guy winning (say Romney taking VA, OH and FL). But the question is how many paths like these lead to your guy winning versus paths that lead to the other guy winning? How likely is the path you see compared to all the other paths?

    Some of the analysts I saw did notice that. They noticed, for example, it would be almost impossible for Romney to win if he lost Ohio but Obama could win if he lost Ohio. Silver, though seems to have been rigerous about building up a huge population of all possibilities including the nearly absurd (the one in a million shot Romney takes CA and NY) as well as more common possibilities (R takes OH, VA, O FL or R->FL, VA, O-> Oh etc.)

    But if you want to just be a ‘cheerleader’ type pundit, it’s easy to pluck out a few results that confirm your bias and run with them sounding very, very smart along the way. The question is do the cheerleaders know they are that, just playing their audience for pay or have they conned themselves into thinking they are smart? Are they crafty people playing the fools or fools playing crafty people? I would say Karl Rove is crafty, Ann Coutler, Dick Morris, etc. are fools who think they are smart.

  24. Except that in past elections where the Dems lost … those on the left were the ones trusting to wishful thinking and gut feelings and …

    That is not true! What makes you think that?

    Apparently Mr Silver is the only pollster in existence. We’re not wrestling details. You claim polling is accurate. I point out many times in which polls are “surprised” and results fall outside of error bounds. When this happens in science, red flags go up. Yet, here it does not. Why is that?

    The reasons I am holding up Mr. Silver are several. First, he specifically was the one that everybody on the left (and center) came to trust because of his record, his methods, and his writing and explanations. Second, he specifically was targeted by FOX and other Republican propaganda outlets as being biased and wrong. Third, he used polls from all over the country, thereby showing it’s not just him and not just one specific poll that works, but that polls in the aggregate work.

    There are many possible reasons that there are surprises, as you should well understand. Individual polls can be flawed in several ways from basic errors to bias to sampling. Also, reality on the ground can change between the time of the poll and the time of the election.

    Nobody is saying you should take every poll as gospel. We are saying that if you look at multiple polls and analyze them carefully, you can get extremely accurate information. That is what Nate Silver did. That is what we knew he was doing AHEAD OF TIME. On this very blog I linked to him and advised you to read his blog if you wanted to understand polling. Your side doubted him and doubted the polls and were wrong on both accounts.

    Tell me this. In the next election cycle in two years, will the published and now well estbalished methods used by Mr Silver be used universally by all polling agencies? If not. Why not.

    This question does not make sense because Mr Silver is not a polling agency and does not do polling.

  25. The question to ask then is would Silver’s method consistently produce accurate results in election after election. A problem on the local level may be a lack of polls. Will there be enough funded to collect the data to do so? If not then the method might be sound but it only will work on big elections.

    As a side note, my method of following the betting markets also did pretty well. You could have purchased Obama contracts for $7.00 or sold Romney contracts for $3.00 as late as Tuesday afternoon and scored an instant 30% return before midnight.

  26. I don’t know how much credibility Mark gave to the ‘skrew’ argument but here is the actual results here:
    http://www.slate.com/blogs/moneybox/2012/11/07/unskewed_election_results_what_if_you_re_weight_for_party_id.html

    The skew argument went something like this. 40% of people are Repubicans, 40% are Democrats and 20% are neither. If you do a sample and get 45% Democrats, 35% Republicans and 20% neither, you should downplay your Dems and upplay your Republicans. I previously explored why this is done with more unchanging traits like gender….if you get a sample of 75% men you shouldn’t just assume 75% of voters will be men.

    Well it seems like the skew was real. More voters identified themselves as Democrats than Republicans….which makes sense. If a person is going to vote for the Democratic candidate, they are more likely to say they are Democrats…even if more ‘official’ lists do not have them as such.

  27. JA,
    It seems in 2010 Mr Silver predicted that the Dems would keep the house. In your effulgent praise of Mr Silver and the rock solid basis of his methods it is odd that you would omit that detail.

    That is not true! What makes you think that?

    It is not? What pundits on the Democrat side predicted a Kerry defeat, your thesis would suggest that all but a few would do so.

    That is what Nate Silver did. That is what we knew he was doing AHEAD OF TIME.

    So. Explain 2010.

    This question does not make sense because Mr Silver is not a polling agency and does not do polling.

    He has methods. It seems any polling agency could do the same to report their results, i.e., aggregate their and other polling agency raw results the same way in their reporting. We have one number for the mass of an electron. Not an “what Gallup says” vs “what CBS got” &c. Again, results which like outside the error bars need explanation, this is not true with polls. You keep dodging this. Explain why.

    And as for “not being a polling agency” … who writes his checks and why?
    Boonton,

    The question to ask then is would Silver’s method consistently produce accurate results in election after election.

    Note the above. It would seem not.

  28. Mark,

    Perhaps you can explain your electon analogy a bit more…is the mass of an electron determined by fundamental laws of physics (like the value of pi) or is it a figure we have from observation and measurement?

  29. It seems in 2010 Mr Silver predicted that the Dems would keep the house. In your effulgent praise of Mr Silver and the rock solid basis of his methods it is odd that you would omit that detail.

    According to wikipedia the result was within the confidence interval.

    What pundits on the Democrat side predicted a Kerry defeat, your thesis would suggest that all but a few would do so.

    I don’t have that data, just my memory, admittedly unreliable (although yours is too.) If you have the data, I’d be happy to see it.

    Again, results which like outside the error bars need explanation, this is not true with polls. You keep dodging this. Explain why.

    How am I dodging it? I’ve said repeatedly that individual polls can be flawed, sometimes intentionally (Rasmussen, etc.) If you mess up your methodology, your error bars are worthless.

  30. Indeed, he projected a gain of 54 seats where the actual was 63. He was off by 9 which is just 2% of a population of 435 Representatives.

    Notice that this contradicts Mark’s point, a liberal pundit predicting a Republican win rather than simply always predicting liberal wins. Silver was slightly off on the 2010 prediction but he wasn’t off in predicting a Republican gain.

  31. Boonton,
    I saw that he predicted a Dem holding of the house.

  32. JA,

    sometimes intentionally (Rasmussen, etc.)

    In 2008, of the polling agencies Rasmussen was closest to getting it right. Woops.

    Boonton,
    The quote on Mr Silver that I saw may have been based on a slightly earlier than the 2-days before prediction quoted by wiki.

  33. I saw that he predicted a Dem holding of the house

    The Republicans had 179 seats before the election so if Silver had predicted a 54 seat gain he would have also predicted GOP takeover. The wikipedia source you cited quotes Silver directly on the 54 seat gain so I think that would settle it.

    I think one aspect of this argument that is interesting is Silver doesn’t just average polls together. He subjects them to a ‘half-life’ so their weights fall off if the poll was held long before an election. He also weights them based on their previous accuracies. So if one poll is consistently less accurate than another, it’s weight in the average gets reduced accordingly.

    The description does not say that Silver weights polls per bias. For example, Intrade isn’t a poll but it has had a consistently and somewhat mysterious Republican bias (it’s not clear whether that’s caused by right-leaning people purposefully blowing money on Intrade to create positive buzz for Republicans or maybe it’s just that prediction markets are not yet deep enough to arbitrage out these ‘bumps’). But it seems like if such bias was there, as the right alleged, it would be a simple thing to pick up and accomodate. If, say, Rasmussen, consistently gives Democratis a 1 point edge it would be very easy for someone like Silver to pick up on that and knock Rausmussen’s Democratic figure down 1 point in his model.

  34. Mark,

    The last poll in 2008 seems like a real outlier in accuracy for Rasmussen. In general, they have a consistent and predictable Republican bias.

  35. Boonton

    Notice that this contradicts Mark’s point, a liberal pundit predicting a Republican win rather than simply always predicting liberal wins.

    I thought you said Mr Silver was a statistician … now he’s a liberal pundit. Does he write political commentary? Books? What? What qualifies him as a liberal pundit?

  36. I’ve seen him described as an Obama supporter and he got his start as a diariest on the Daily Kos. I haven’t read him enough to know if he goes into politics beyond simply analyzing numbers, though.

  37. Might want to review this short but interesting piece on the right wing media bubble

    http://www.theatlantic.com/politics/archive/2012/11/how-conservative-media-lost-to-the-msm-and-failed-the-rank-and-file/264855/#.UJqSwlJ-XZI.twitter

    I’d be curious to know your thoughts.

  38. JA & Boonton,
    Is something else going on here? You’re adamantly defending Mr Silver, saying that he was attacked viciously by right wing bloggers and pundits. Uhm, I never did. I didn’t even know it was happening. Yet, apparently I’m accountable for it.

    Apparently you figure polls are almost useless, agreeing with me, except that certain people with judicious use of poll aggregation can produce prediction. Soon, we’ll be able to do away with the elections entirely. Yet, the problem remains that polls exist frequently which offer prediction whose error bars don’t include the actual results. You can’t stuff this under the rug by saying “aggregation” will correct that. It’s not how it works. If your error doesn’t include the result .. your result is wrong (or the you have a new discovery).

    Boonton,

    I think one aspect of this argument that is interesting is Silver doesn’t just average polls together. He subjects them to a ‘half-life’ so their weights fall off if the poll was held long before an election. He also weights them based on their previous accuracies. So if one poll is consistently less accurate than another, it’s weight in the average gets reduced accordingly.

    I took the liberty of highlighting a sentence there. How do you measure “previous” accuracy? You mean from 2,4,6,8,10 years ago, i.e., the previous elections? Or by some other measure?

    An interesting question might be why those polls which he gives lower weights … have lower weights in the first place.

  39. Apparently you figure polls are almost useless, agreeing with me, except that certain people with judicious use of poll aggregation can produce prediction. Soon, we’ll be able to do away with the elections entirely

    I don’t believe this follows. Mr Silver could encounter an election that is truely 50-50. I suspect, for example, the 2000 election would have been like that if he had been applying his model then. In that case he would only be able to tell us that the result would be a coin flip. That’s useful information but not sufficient to just not do the election and pretend we did. Even this election he had a 90% confidence meaning that if 10 elections happen along these lines at least 1 time we can expect his model to call it wrong.

    And as for getting rid of elections, I think this confuses the act from the result. It’s helpful to know a drunk driver will be pulling out of the bar and crossing main street at 7 PM if you’re a cop. That doesn’t mean you can arrest the drunk driver before he leaves the bar. It was a sure thing Reagan would trounce Mondale in 1984, yet that doesn’t mean it wouldn’t have been an outrage if ‘philosopher kings’ had declared they wouldn’t bother with an election in 1984. The act of voting is a ‘buy in’ for the population. Simple polls, IMO, are not a substitute for that.

    I took the liberty of highlighting a sentence there. How do you measure “previous” accuracy? You mean from 2,4,6,8,10 years ago, i.e., the previous elections? Or by some other measure?

    Poll X’s average error over the last election was 3 points. Poll Y was 1 point. Hence Y should get weighted 3 times X’s weight. I suggest you dig into his actual methods but in theory it doesn’t seem like that difficult an exercise.

    An interesting question might be why those polls which he gives lower weights … have lower weights in the first place.

    I think some of the answers may be pretty mundane. Some polls take smaller samples, for example. Bigger samples will be more accurate but will cost a customer more to fund. What’s interesting IMO isn’t that some polls are less accurate over time, but that he doesn’t see much value in building a bias compensation into his system. He doesn’t seem to find that polls on average are biased for Dems hence no need to compensate for that.

  40. Mark,

    I am holding you accountable for two things: the “cricket races” thing is the more specific, but in a way stands in for the way that you find excuses to disbelieve whatever avenues of ascertaining reality that educated non-right-wingers use to… ascertain reality in fields as disparate as elections, global warming, religion, and the economy.

    I’m talking about Nate Silver because he has come to represent polling this cycle. Maybe it wasn’t on your radar, but there was a whole thing where the right was loudly dismissing him — and by proxy polling in general — because he was saying something they didn’t want to hear. You may not have done it personally, but again, it falls into that same pattern of dismissing what unbiased people take as the obviously best (not necessarily perfect, just best) way of ascertaining reality in favor of gut feelings, wishful thinking, and biased pundits.

  41. JA,

    I’m talking about Nate Silver because he has come to represent polling this cycle. Maybe it wasn’t on your radar, but there was a whole thing where the right was loudly dismissing him — and by proxy polling in general — because he was saying something they didn’t want to hear

    I’ve regarded polling as an inaccurate device for far longer than just this cycle. I don’t recall every thinking opinion polling was worth much salt, well, ever. It isn’t a “this cycle” thing. So your second complaint is misguided. This isn’t a “this cycle” thing.

    Your regarding as polling as the “obviously best” method of ascertaining reality … is weird … you don’t even do it yourself. It’s akin to thinking that “the customer is always right” is a good maxim for business … except that what the customer wants (if you poll him) is perfectly wonderful product for free. Remember, If you poll Spaniards vs Irish on whether they are religious you get a 90%/10% split between the two countries but if you look into whether they worship weekly you get a 10%/90% split. How useful is your poll? Not very. And that is exactly the problem. People rarely act in the way they claim in a poll. That is the fundamental problem with polls. But it’s not the accurate thing you pretend rhetorically that it is, whether or not you believe what you claim is another matter. Which is why I keep pressing you on results which differ outside of their claimed error. In real hard science such a thing cannot happen. It’s impossible. If you measure the mass of the electron at .6 +/- .05 Mev …. that is a problem. What it is not is a “result” to be averaged in. Two results which don’t match on error bars are different. Somehow polling is different. And that is a thing for suspicion and distrust not to ignore.

    You say polling is accurate. Except … you don’t. When pressed, its not polling that’s accurate, but instead what is accurate is just one single solitary aggregator of polls. He’s accurate. Nobody else apparently. How? Dunno. His method is secret. Unpublished. It’s not an algorithm. It’s his industrial secret. Ergo, it’s “art” not science. Now to be honest, much of science is art … not science. But you’ve rejected Polanyi out of hand. You don’t think science is art. Yet polling as art … that’s science. Gotta a little cognitive dissonance going on there.

    but in a way stands in for the way that you find excuses to disbelieve whatever avenues of ascertaining reality that educated non-right-wingers use to… ascertain reality in fields as disparate as elections, global warming, religion, and the economy.

    Your issue with “these ways to find excuses” is hypocritical. You make the claim that with religion that you checked into the claims researched and found religious claims wanting. But you have done no such due diligence with any other of these things, yet in many of these fields you hold the certainty with which their results far higher than actual practitioners within the fields themselves. You are more sure of the rightness and correctness of neo-Keynesian economic prescriptions than those within the field itself. Not for nothing is economics called “the dismal science” (I’ll let you google for yourself the origin of that title). You claim climate science is on firm footing. Yet cannot name a single freaking prediction that the field has staked out that has turned out. People who do computer simulations for a living often refer to simulations as computer aided story telling. But you don’t see it that way. You give more credence to their results than the practitioners do. String theory is widely dismissed as “mostly useless” in Physics because it has made no predictions in 30 years. It won’t be on firm footing until it can make predictions and demonstrate reality matches that. Apparently however, that is not a claim that this new field of climate science need to have to before you think governments (not individuals like yourself mind you) need to devote many trillions of dollars to act on NOW NOW NOW or humanity is doomed. Apparently however, a query for results is “an excuse” to disbelieve. You reject religion because to your mind it lacks results. You do not reject climate science on the same grounds. Why? I suspect it is merely because its clerics wear different garments and use computers … or perhaps its a team thing, those priests are liberals like yourself, so you “know” they won’t deceive you.

  42. Boonton,

    The act of voting is a ‘buy in’ for the population.

    Ah. So you agree with my description Tuesday of voting as a liturgical act, eh?

  43. Your regarding as polling as the “obviously best” method of ascertaining reality … is weird … you don’t even do it yourself.

    I think polling is clearly the bet way for ascertaining what will happen in an election short of actually waiting till the election happens and seeing what the results are. You are free to nominate a better method and demonstrate its superority by pointing to lower variances between predictions and results. A huge amount of money spent by political advisors on polling might be captured by such a better method.

    There’s two possible explanations for Silver’s tract record. One is that he just happens to be really lucky and has fooled himself into thinking systematically averaging multiple polls is a system….kind of like the bingo player who wins twice in a row while wearing a green outfit believes wearing green makes one win at bingo…..or using polling data intelligently is an accurate way to predict election results. The fact that this does not produce strict determinism (being able to predict results exactly all the time) is not sufficient to dismiss it as an ‘art’ rather than science. Silver’s methods of averaging and assigning different weights to polls either contribute towards making more accurate predictions or detract from making more accurate predictions. That can be objectively measured just as we can measure whether a certain shaped part has a higher or lower failure rate than some other shaped part in a machine.

    Ah. So you agree with my description Tuesday of voting as a liturgical act, eh?

    I suppose, I guess it’s like putting your signature on a contract. The actual ‘buy in’ and support of the deal happens much earlier, but the act of signing itself boils it down to a single act forcing the person doing it to either say yes or no in one decisive act. I’m not sure liturgical acts are exactly in that mode since one can to go mass and ‘go through the motions’ without really meaning it. You don’t vote or sign a contract ‘without meaning it’. By voting it counts and by singing you bind yourself to the deal, the opinion or hunch becomes a concrete act. I guess some liturgical acts like getting married or perhaps taking communion are in this form…you are free to not do them but to do them means you must actually do something. You tell me how complete the analogy really is….

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>