Tuesday may turn out to be a nightmare

I have reanalyzed the electoral map with the latest information. Applying some reasonable criteria (see below) to the most current information, a very reasonable conclusion is that the electoral vote not counting Pennsylvania will be Obama 252 vs. McCain 265. Neither of these numbers is above the required number to win. Starting at this base, the candidate who wins Pennsylvania wins the election.

I am not prepared to put Pennsylvania in either column for the following reasons: 1) All of the effects in my adjustment criteria (outlined below) are strongest in Pennsylvania and 2) McCain and the Republicans are putting everything they have into Pennsylvania. The numbers show that this strategy is working. No reasonable observer can claim that this strategy can’t work. It is a matter of probabilities and a McCain win in Pennsylvania is a possibility.

Here is what my new, Nightmare Map looks like:
i-9fcab902e17c2fcf650f3e3c08d181dd-the_nightmare_scenario_election_2008.jpg

The criteria I used to make this map:

Using data, as well as the map itself, from Real Clear Politics:

1) I assume the Bradley Factor, the Bubba Factor, the Walmart Women Factor, etc. among undecided voters as well as previously ‘committed’ voters to be four percent. Taking the best estimate available, an Obama lead of four percent or less equals a McCain win.

2) Using non-partisan polls, if recent polls show a shift towards McCain and there are a lot of undecideds in the poll, I assume that this indicates a real shift among undecideds towards McCain. In this case, I use that latest post shift number to estimate the spread, rather than the RCP spread. I don’t need to come up with a number. I just need to guess if the current trend is going to put the difference to four points or fewer.

3) States that are obviously strong Obama or strong McCain are, obviously, moved into the appropriate column.

Strictly applying these criteria to Pennsylvania, Obama still wins, but that is today. The situation has shifted from Friday to today. With the extra effort being spent by the Republicans in Pennsylvania, the situation could continue to shift between now and Tuesday, and this shift could be at a higher rate than the shift we are seeing in other states.

I just listened to an extensive interview with a reporter who has been covering Ohio, and everything he is saying is suggesting that Ohio will break for McCain. State by state, this sort of thing is happening.

By the way, if the cable news does not drop this Palin Pranked story this is going to be seen as an unfair attack on her and will give McCain Palin an extra point across the board. As a Minnesotan, I can tell you that the backlash effect can be very very painful (this is how we got Norm Coleman as Senator).

Also, by the way, the national press is going to have to stop congratulating the McCain campaign for not dragging Jeremiah Wright out in this election. They are doing this now.

Comments

  1. #1 penn
    November 2, 2008

    Is this a joke, or just the absolute best possible scenario for John McCain?

  2. #2 kraut
    November 2, 2008

    America deserves another term being run by neocon ideology, encompassing all the follies having been perpetrated during the first set.
    A people that do in their majority not realize what has been done in their name to themselves in their country – stripping away rights and opening the way for a form of fascism “made in USA” – and to victims of their politics in other countries, deserve no better.
    A people who can conceive of another gig by those that created the mess – politically, economically – the country finds itself in, deserve what they get.

    Maybe after a second set by the republican party, the demarcation lines between democracy and an overtly fascist state (fascism is the complete meshing of capital and politics, as communism is the complete meshing of labour and politics)will have been clearly laid out – but it might then be too late to do anything about it until a total collapse occurs, as happened in europe.

  3. #3 Dave C
    November 2, 2008

    Please don’t quit your day job!

  4. #4 MPW
    November 2, 2008

    Greg, your SciBling Mark Chu-Carroll over at “Good Math, Bad Math” has an interesting post about misunderstanding and misreporting of polls in the media. He clued me into FiveThirtyEight.com, “a site where a guy [Nate Silver] who is very clued to statistics tries to combine results from all of the available polling, including factoring in weights based on the sampling methods used by the various polls. I think his method looks very good, and that he’s likely the most accurate predictor. But we won’t know for certain until his method is put to the test, by seeing how well it really matches the final results next week.”

    FiveThirtyEight is confidently predicting an Obama win with 340-something electoral votes. There are a few posts on the front page now about why Pennsylvania almost certainly isn’t an issue.

    Also relevantly, Silver is highly critical of RealClearPolitics, claiming their sampling methods tend to over-favor the Republicans (he even accuses them of doing this deliberately for partisan reasons). It’s a fairly convincing argument.

    The “Bradley Effect” has come under a lot of scrutiny lately from people who think it isn’t a factor here – that it was perhaps an illusion to begin with, or that it’s lessened or vanished now, or that Obama has largely managed to escape it. Some think he might even benefit from a “reverse Bradley effect” where some voters in areas with strong bias against voting for a black candidate are planning to vote for Obama (in large degree for economic reasons) but are keeping it under their hats (John Deere caps?)

    As far as I understand things, if you average out the polls, Obama has been over 50% and up 6 or 7 points all week. This is barely budging, and there’s basically one news day left until election day. Silver has been reporting, as few in the media have, on the McCain campaign’s virtual abandonment of the Republicans’ vaunted get-out-the-vote ground game – an area where the Obama campaign is famously tireless and innovative.

    I’m not saying you’re off your rocker or that what you’re positing is impossible – just that it seems highly unlikely. But after the last two elections, Democratic supporters are certainly justified in being skittish and a little paranoid.

  5. #5 The Ridger
    November 2, 2008

    Electoral Vote.com also calls it a landslide for Obama, based on the eight big polls.

  6. #6 Zeno
    November 2, 2008

    Frankly, I’m glad that Greg did this. No, I don’t really believe this is going to occur, but Greg has gone to the trouble of trying to put some flesh on the bare bones of the “nightmare” scenario, which might be better labeled as “McCain’s pipe dream” scenario. The bright red map above (despite the well-known distortions of the big empty red spaces of the country) is chillingly effective at shaking people out of any complacency they might have. Is it going to happen? No, no, no, and no. Provided we don’t slack off and we deliver the Obama vote.

    Thanks for going “Boo!”, Greg.

  7. #7 Greg Laden
    November 2, 2008

    I’m well aware of not only five thirty eight, but I’m intimately knowledgeable of the statistical techniques they are using, and I approve. There is no probelem any of that.

    But all of the statistical models make as set of assumptions that are distinctly different from the assumptions I’m making. You can’t disagree with my map because someone else has a different map. You have to disagree with my map because you think my assumpitons are not as good as the assumptions underlying the other studies.

    Penn: Yes, this is the best possible McCain scenario, and entirely within the range of possibilities.

  8. #8 yogi-one
    November 2, 2008

    Point being: get out there and do your civic duty. Nothing can be taken granted.

  9. #9 Pierce R. Butler
    November 2, 2008

    … this is the best possible McCain scenario …

    Would it be better for McC to become a footnote in the history books, or to earn full chapters as the president who, completing the destruction wrought by his predecessor, was worst than the worst*?

    (*until his rubble-bouncing successor took office mid-term)

  10. #10 becca
    November 2, 2008

    Ah but you’re forgetting the CareBear vote.
    Relatively young people with no land line who have never voted before are unlikely to be in the polls. However they are going to break overwhelmingly for Obama.

  11. #11 Greg Laden
    November 2, 2008

    Becca:

    Here is what I know about this:

    1) Everybody mentions it all the time. Including me until about a week ago when I looked into it.

    2) The major polling companies claim that they know about this and compensate for it. They do call cell phones. (They don’t use phone books to make these calls. They generate random numbers and that means they can include cell phones easily)

    3) There is one study, from St. Cloud state, that made a special effort to adjust for this even more, that came up with the wackiest results I’ve ever seen in a poll, and which showed, if it is correct, that the majority of people with these cell-phones are voting for Coleman here in Minnesota, not Franken, and they tend overall to be republicans.

    The young turks with the cell phones only are not liberals, that study tells us (if it is valid).

  12. #12 hgordon
    November 2, 2008

    Excellent post. You have captured the essence of this election, which is “hope” vs “fear”.

    It is difficult to gain much confidence from watching the pollster projections bouncing all over the map, but I think it will come down to the effectiveness of the ground campaigns.

    Along those lines, there is an interesting article in the Christian Science Monitor – see “My wife made me canvas for Obama; here’s what I learned” – http://www.csmonitor.com/2008/1103/p09s02-coop.html . We really won’t know the real story until the votes are counted, but this kind of story gives me “hope”.

  13. #13 Stephanie Z
    November 2, 2008

    hgordon, um, wow. Thanks.

  14. #14 Lynn
    November 2, 2008

    hgordon,
    wow from me, too :)

  15. #15 Shadron
    November 2, 2008

    Lord, guys! Go consult someone who knows, perhaps: http://election.princeton.edu/, the Princeton Voting Consortium.

  16. #16 Lycosid
    November 2, 2008

    PA will go Obama. I live about 20 minutes from Dover, PA and about 5 from a union building that was spraypainted with the word “Communists” for displaying an Obama sign; but I can tell you that Philly and Pittsburgh decide any elections here and they’re both solidly Democratic.

    Also, not that we’ve had any other choice, but any state where a mob-owned pig like Rendell can keep being elected governor is going to vote blue.

  17. #17 R. A. Fisher
    November 2, 2008

    This is unfounded fear mongering. For analyses done by experts, see http://www.FiveThirtyEight.com

    They project a 94% chance that Obama will win in the Electoral College.

  18. #18 greg laden
    November 2, 2008

    Fisher, you are a faker. The real RA Fisher would have read my comments above and learned quickly how foolish this comment of yours really is.

  19. #19 quantum electiondynamics
    November 2, 2008

    I agree with Greg here. Electoral projections treat different swing states as statistically independent – but this is unjustified. Pundits say McCain is lost because “he’d have to win ALL of these states – which is a pretty big coincidence”, but it’s not, if there’s systemic effects that correlate their outcomes. The Bradley effect is an example, but we really should be thinking about the general case: systemic effects like poll biases can cause correlated swings in votes, which makes the total variance much larger than predicted by independent variables. (With independent variables, uncertainty decreases with increasing N – it goes as 1/sqrt(N).)

    I believe “electoral distribution” simulations on 538 are an egregious example of this error.

  20. #20 David Marjanovi?
    November 3, 2008

    - No, most of the big pollsters don’t call cell phones.
    – The Bradley Effect probably didn’t exist to begin with; in the last poll in that election, Bradley’s lead had shrunk to 1 %.
    – Even the pollsters with a consistent Republican bias (relative to all other polls), such as Rasmussen, Mason-Dixon and of course Strategic Vision, find PA for Obama by a large margin.
    – In the polls of 2004, Ohio was a dead heat, swinging back and forth daily, and so were Iowa and Pennsylvania. In this year’s polls, Obama has a moderate but consistent lead in OH, every day, and he’s got IA and PA in the bag.
    – Doesn’t the fact that states like VA, NC, IN, MO, ND, MT, GA, AZ are in play at all tell you something? Or, on the other side, do you remember how McCain wanted to believe for months that Michigan was a swing state? Well, in the 2004 polls it was. This year it’s a done deal, and has been for months. Even during the convention bounce with national drooling over Palin not one poll ever found it Reptilian. Same for Iowa for that matter.
    – Even some states that are still in the bag for McCain will vote for him by considerably smaller margins than they did for Bush in 2004. Utah was 26-71 Kerry-Bush (and 27-67 Gore-Bush in 2000); now it’s 32-56 Obama-McCain. Compare Clinton’s reelection: 33-54 Clinton-Dole. In a lot of states this year’s polls look like Clinton’s reelection. South Dakota, for example: 1996 43-47, 2000 and 2004 38-60, now 44-53. If all the factors you mention that could introduce a hidden bias for McCain, why didn’t they introduce the same bias for Bush and Dole?

    Yes, go vote. But once that is done, I don’t think they can steal that much in that many states without it being just too obvious. Also, Kenneth “Katherine” Blackwell, who counted the votes of Ohio in 2004 on his desktop, has been replaced by a Democrat.

  21. #21 David Marjanovi?
    November 3, 2008

    Oops, incomplete sentence:

    If all the factors you mention that could introduce a hidden bias for McCain

    really exist this time, surely they were at least as strong previously — so why were they so much smaller before in the last three elections

    , why didn’t they introduce the same bias for Bush and Dole?

  22. #22 Minnesota Central
    November 3, 2008

    Good news … you’ll find out early tomorrow evening that all your concern is for naught.

    McCain is in PA for two reasons :
    #1. PA does not have early voting. VA, NC and FL are already started and Obama is banking votes. PA can still be affected.
    #2. PA has an lot of older voters who McCain has appeal … especially veterans … and Pro-Lifers.
    Now, let’s remember that PA voted for Pro-Choice Spector and liberal Casey and Kerry. Is there any reason, why they won’t vote for Obama ? McCain thinks because Clinton beat Obama, that those voters will be his … wrong … there was not a bit of difference from liberal Clinton and liberal Obama … it’s a different race now that it is race a liberal versus a conservative.

    Lastly, you are giving every state to McCain … it’s just not realistic.

  23. #23 Greg Laden
    November 3, 2008

    David: All good points, and you might be right on every count. But…

    I’m not sure if you are correct about the cell phones. Random number generation should call cell phones, and I swear, I saw some guy who was some kind of expert contradicting you on TV. But you could be right. I’m just not going to base anything yet on hearsay which is all I have so far, other than the St. Cloud State poll.

    I agree that the bradly effect didn’t exist in the case of the actual Bradly, but there is a distinct possibility of it existing in a national election of this type.

    Mason Dixon’s margin is almost close enough to be MOE. They are two points outside the MOE. If there are two points of ‘effect’ biasing as suggested by me in this post, the national race is at present dead even according to Mason Dixon.

    The swinging related to 2004 vs now is not relevant to a hidden consistent even if small biaes. The whole point of this post is that this is a different race.

    Doesn’t the fact that states like VA, NC, IN, MO, ND, MT, GA, AZ are in play at all tell you something?

    Yes it is very meaningful. Not really related to the issue at hand. An analysis that uses “map redrawing” as a basis would be an entirely different kind of analysis, and would necessarily be based on zero prior knowledge.

    But yes, it is interesting and meaningful.

    Keep in mind, I do have more than one post on this topic … with different conclusions.

    No matter what, if you are a democrat, not only go and vote, but also, take Tuesday off and join a GOTV effort!!!!!

  24. #24 Greg Laden
    November 3, 2008

    MC: Good point about PA vs FL, etc. Had not thought of that.

    As far as voting patterns in PA: This is entirely a matter of turnout. Anti-choice voters did not vote for pro-choice candidates. They are just not as organized.

    As for your last point, you are young. Or you would not say that! Ever hear of richard nixon?

    But I hope you are right. But just in case, VOTE.

  25. #25 David Marjanovi?
    November 3, 2008

    I agree that the bradly effect didn’t exist in the case of the actual Bradly, but there is a distinct possibility of it existing in a national election of this type.

    Fine, but if so, you also have to take the Reverse Bradley Effect into account: people in racist areas telling pollsters they won’t vote for the black guy, when in fact they will. I remember reading that someone analyzed the Democratic primaries and found the Bradley effect in some states but the reverse Bradley effect in even more.

    No matter what, if you are a democrat, not only go and vote, but also, take Tuesday off and join a GOTV effort!!!!!

    You bet I would if I were a US citizen or (concerning the stuff behind the last comma) even just lived on the same continent…

    ————————-

    As I alluded to, I don’t disagree that tomorrow may turn out to be a nightmare. That’s because of the ballotless voting machines in PA, GA and a few other states, especially those (like GA) where the election is ultimately managed by a Republican. Even with ballots, massive fraud is possible if you can get away with not recounting them, so massive that the outcome changes — see Ohio 2004, which Kerry won except for the counting. But, as I also said, getting away with stealing a landslide is another matter, and Ohio 2004 was not a landslide for anyone.

    At least, Bush v Gore contains a clause that forbids it from being used as a precedent for anything.

  26. #26 Irvin Schonbrun
    November 3, 2008

    Maybe the number of Bush-family-related or -bribed state governors might also be a factor?

    Who says it can’t happen here?

  27. #27 Pierce R. Butler
    November 4, 2008

    According to http://www.electoral-vote.com/evp2008/Info/polling-faq.html#cell

    Do pollsters call cell phones?

    No. It is illegal for them to do so. … Since these people tend to be mostly young people, the pollsters intentionally overweight the 18-30 year olds to compensate for this effect…

    Calling a “randomly-generated number” ain’t much of a loophole.

    If my experience with a dozen hang-ups/day here in Florida is any guide, they don’t poll any of us who screen our calls, either.

  28. #28 Pierce R. Butler
    November 4, 2008

    Oopsy damnit – the above should be formatted thusly:

    Do pollsters call cell phones?

    No. It is illegal for them to do so. … Since these people tend to be mostly young people, the pollsters intentionally overweight the 18-30 year olds to compensate for this effect…

    But, on second reading, this explains why the younger generation is tending to obesity – somebody’s got to stop those pollsters!

  29. #29 Greg Laden
    November 4, 2008

    So, the point is that polling adjusts for the cell phone effect, so we should not expect to see a cell phone effect.

    Unless there is an unestimated extra cell phone effect of course. It still concerns me taht the one study that supposedly did this showed the opposi

  30. #30 Becca
    November 4, 2008

    I think the way they add the adjustment for the cell phones might not be helpful. Cell phones are more common among younger folks; however I’d also be shocked if they weren’t more common among doctors and richer folks. I view it more as a multipurpose “some phone numbers are more likely to be called than others” confounder.
    Pierce- thanks for the FAQ, it was very handy. I’m especially curious about the effect of Vonage type phone numbers… anyone calling my 708 number is going to expect an Obama supporter. On the other hand, I tend to screen my 717 landline a lot more, so they’re less likely to get any answer.
    Also, just to clarify CareBear is only “relatively” young (i.e. not in the 18-30 set). I wonder how many other African American first time voters there will be… in the 40+ age group.
    (and this is a bad thought I’m ashamed of, but it’s more a reflection on the evils of our society than my personal evilness- convicted felons can vote in Pa, so we might get more first time male African Americans than we would otherwise).

    But Greg, are the polls you used of “registered” voters, or “likely” voters? One of my points with the ‘CareBear vote’ is that he would in no way show up on a screen of “likely” voters.
    Of course, I’d be completely unsuprised if CareBear would refuse to talk to a pollster on the grounds that he doesn’t want to give out any more information than he has to. I’m still trying to figure out if “paranoia” is higher in one population than another- there’s so much rich data to analyze.

  31. #31 majolo
    November 4, 2008

    The quote from Pollster.com above is incorrect. Several polls do include cell phone samples, as Pollster itself noted back in May at http://www.pollster.com/blogs/more_cell_phone_survey_news.php
    The law apparently prohibits use of an autodialer and/or recorded messages when calling cell phones, so I would guess polls use a randomly generated list which must be dialed by hand. There is a good roundup and comparison of polls with and without cell phones in this election at http://www.fivethirtyeight.com/2008/11/cellphone-effect-continued.html

  32. #32 Greg Laden
    November 4, 2008

    OK, we’re going to have to take a poll: How many people think polls use cell phones, and how many people think polls don’t use cell phones?????

    Show of hands, please.