Skip to main content

There have been a lot of articles in the media lately saying that 538 is overestimating Obama's chances.  I don't blame them.  Republicans need to keep their supporters enthusiastic.  If it is a forgone conclusion that they will lose, who would want to spend their time volunteering or give money to the campaign? Those in the traditional media want an interesting story.  It's in their financial interest to make this campaign seem as close as possible.

Nate Silver currently estimates the president's chances of winning at 83.7%.  I'm going to make the case that this is too LOW.  If I had to guess, I would put the president's chances at closer to 95% right now.  Here's why:

1) Lag time
Nate Silver has admitted himself that his model is slow to react to trends.  The model is mostly based on polls.  Some of the polls included in averages started in the field 5 or 6 days ago.  President Obama's widely praised response to Hurricane Sandy will take a while to digest.  If this were a non-election week, it might take until next Wednesday to fully see the extent of Obama's bounce.

538 also smooths out short-term fluctuations.  It is slow to react to poll changes when there is a reason for the fluctuations.  It took over a week for the model to really show the effect of the Denver debate.   I expect Obama's win percentage to continue to creep upward as we get closer to Tuesday, but right now, 83.7% is too low.

2) GOTV
Most anecdotal evidence points to a vastly superior Democratic ground game.  How do you quantify this?  You can count campaign offices or campaign contacts, but this is very crude and probably totally inaccurate.  Consequently, the model can't take this into account.  How could it?

This year has shown a very high disparity between registered voters and likely voters.  It has been around 5-6% as opposed to the 1-2% that we see in normal years.    It is my theory that much of this discrepancy has to do with Hillary vs. Obama in 2008.  Both candidates made extraordinary efforts to register voters in all 50 states.  Many of these voters in non-swing states might not bother to vote this time.  However, in swing states, there will be a strong GOTV effort, I just can't imagine a 5-6% gap.

3) The MATH
Barack Obama has 4 distinct paths to victory:
Win OHIO
Win FLORIDA
Win VIRGINIA
Win COLORADO
I'm going to give the president Wisconsin.  His lead there is comparable to his lead in Pennsylvania, and I just don't see him losing it.  The president has fairly significant leads in all of the other swing states except North Carolina.  If he wins NC, he also wins Virginia and therefore the election.

If Obama wins Florida, it's over. If he wins Ohio, he just needs one of New Hampshire, Iowa or Nevada to win.  If he wins Virginia, he needs two little swing states
If he wins Colorado but loses all the other big states, he needs all 3 little states.
President Obama is a fairly strong favorite to win all 3 little states, especially Nevada

The president has a solid polling lead in Ohio, small leads in Virginia and Colorado, and is about even in Florida.  The only way that he can lose is if all of the state polls are systematically wrong.  Here's the problem:  for Obama to lose, the state polls must be all wrong THE SAME WAY.  All four states are very, very different and Obama has different strengths in each state.  Ohio has the auto sector,  Colorado a large Latino population and  Virginia has a large government population and Florida - well, Florida is Florida.  It's possible but unlikely for the state polls to be systematically wrong in one state.  It is virtually impossible for the state polls to be systematically wrong in ALL four states.  

Let's assume that the polls are systematically wrong in Ohio (a fairly unlikely but plausible proposition).  If the polls are not systematically wrong in the other states, Obama still has about a 95% chance of winning one of the other big states and therefore an over 90% chance of winning the election.

I don't think the polls overestimate Obama's support in Ohio.   Up until a couple weeks ago, The Cook Political report used to publish a characterization state races based on more accurate internal polls.  In all cases, the public polls UNDERestimated Obama's support as compared to his support in public polls.

So take heart everyone.  Work like crazy this weekend.  GOTV.  Democrats are going to win this thing.   If I had a mustache, I would be willing to bet it on the election!

EMAIL TO A FRIEND X
Your Email has been sent.
You must add at least one tag to this diary before publishing it.

Add keywords that describe this diary. Separate multiple keywords with commas.
Tagging tips - Search For Tags - Browse For Tags

?

More Tagging tips:

A tag is a way to search for this diary. If someone is searching for "Barack Obama," is this a diary they'd be trying to find?

Use a person's full name, without any title. Senator Obama may become President Obama, and Michelle Obama might run for office.

If your diary covers an election or elected official, use election tags, which are generally the state abbreviation followed by the office. CA-01 is the first district House seat. CA-Sen covers both senate races. NY-GOV covers the New York governor's race.

Tags do not compound: that is, "education reform" is a completely different tag from "education". A tag like "reform" alone is probably not meaningful.

Consider if one or more of these tags fits your diary: Civil Rights, Community, Congress, Culture, Economy, Education, Elections, Energy, Environment, Health Care, International, Labor, Law, Media, Meta, National Security, Science, Transportation, or White House. If your diary is specific to a state, consider adding the state (California, Texas, etc). Keep in mind, though, that there are many wonderful and important diaries that don't fit in any of these tags. Don't worry if yours doesn't.

You can add a private note to this diary when hotlisting it:
Are you sure you want to remove this diary from your hotlist?
Are you sure you want to remove your recommendation? You can only recommend a diary once, so you will not be able to re-recommend it afterwards.
Rescue this diary, and add a note:
Are you sure you want to remove this diary from Rescue?
Choose where to republish this diary. The diary will be added to the queue for that group. Publish it from the queue to make it appear.

You must be a member of a group to use this feature.

Add a quick update to your diary without changing the diary itself:
Are you sure you want to remove this diary?
(The diary will be removed from the site and returned to your drafts for further editing.)
(The diary will be removed.)
Are you sure you want to save these changes to the published diary?

Comment Preferences

  •  I hope he wins without needing Ohio... (10+ / 0-)

    It might be a good while sorting out their mistakes if it is the deciding state.

    •  Numerous Problems In OH Will Give a Lower Vote (5+ / 0-)

      than the polling. Lots of precincts have been moved, early voting is restricted compared to 08 and 10, there was/is power outage in the blue NE, and there has already been poll watch abuse in some blue areas so more is expectable Tuesday.

      The Dem ground game is great, I don't think we'll lose, but it will be more of a nail biter than it looks today.

      Oh plus the final ad war is brutal. I'm channel flipping regularly from the Buckeye game.

      We are called to speak for the weak, for the voiceless, for victims of our nation and for those it calls enemy.... --ML King "Beyond Vietnam"

      by Gooserock on Sat Nov 03, 2012 at 01:34:01 PM PDT

      [ Parent ]

  •  I agree with your analysis. (10+ / 0-)

    I'll take anything over 270, but I too have a feeling many are going to be surprised Tues, some in a good way, and some in a head exploding way.  

    The fact that Obama is now competitive in Florida is what really does it for me. Don't know if he'll win it, but I believe there's a good chance he can, and that changes this from a close race to a potential electoral blow out, one we didn't believe possible a few weeks ago.  It's all good. Very, very good.

    "A typical vice of American politics is the avoidance of saying anything real on real issues." Theodore Roosevelt.

    by StellaRay on Sat Nov 03, 2012 at 12:57:53 PM PDT

  •  Plus: One Chance in Three of Statistical Bias? (2+ / 0-)
    Recommended by:
    claytonben, RandomNonviolence

    In today's post, Nate said that most of the 16% chance he gives Romney of winning is due to the possibility of statistical bias. He also said this bias could cut either way. I interpret this to mean that he sees about a one in three chance of the polls being inaccurate by 2-3% or more, in one direction or another. This makes no sense to me. One chance in 20? Possibly. One chance in ten? Marginally conceivable. But one chance in three? No way.

    Am I missing something here? If I'm right, this further supports Cosecant's arguments.

    •  I'd like to see you post this on Nate's site. (0+ / 0-)

      Your point seems cogent to me, at least . . . .

    •  Why do you think 1 in 3 (0+ / 0-)

      is too great?

      It seems to me reasonable, given the between-poll variance.

      •  That's sampling error, not bias (0+ / 0-)

        Bias would result from an incorrect likely voter model.

        •  No. Sampling error (0+ / 0-)

          is the "within-poll variance", in effect (or what the "same" poll would give you in many different universes).

          The between-poll variance consists partly of that sampling error, but also, if there is any, of bias.

          Given that the between-poll variance is much more than sampling variance, we know that there is bias (in a statistical, not necessarily a political, sense).

          Given that we know that there is bias, and that it is substantial, and that it seems to have a normal distribution, but not what an unbiased poll looks like, if we make the reasonable assumption that more pollsters are likely to be right than wrong, we can compute the probability that a given poll is biased.  A 1 in 3 probability would be approximately 1 SD from the mean.  So if pollsters who are 1SD in Romney's direction from the mean are showing a Romney win, then Nate's estimate is right.

          I haven't done the calculation myself, but it sounds perfectly plausible.

          What makes you think it isn't?

    •  I Think You're Confusing The *Chance* Of... (0+ / 0-)

      ....statistical bias (which might well be 1 in 10 or 1 in 20 or lower, Nate doesn't say) with the potential effects of said statistical bias - which could be HUGE, depending on the size of the bias.   The 16% is measuring the combination of the two.  

      Think of it this way....the chances of a systemic bias in the polling of a 5 pt magnitude in Obama's favor is very, very small.  But if such a bias did exist then the effect would be enormous.....Romney would be an big favorite to win the actual vote if the polls are systemically understating his support by 5%.

      Nate's 16% is (essetially) approximating the very small chance times the (potentially) very big effect  - and doing so over a wide range of possible bias values.  

      Does that help?

      •  It Doesn't Resolve the Issue (0+ / 0-)

        Nate seems to be saying that there's a one in three chance of the statistical bias being so great that it reverse the likely outcome of the election. In other words, probability and effect are already built into his number. And it's, I dunno, 20% or better. (That's assuming 10% is for bias in Romney's direction and the balance is for days till the election.)

        •  Sounds Reasonable To Me... (0+ / 0-)

          ....even a 2 point bias (and those have been seen before) would make this pretty close to a 50-50 race, maybe 60-40.   I think 16% is at least a plausible number for the bias risk over all values.

    •  I don't really see how you put odds on that (0+ / 0-)

      Statistics can't tell you how likely it is that pollsters likely voter models are wrong.

  •  I think we're in good shape going into the (3+ / 0-)

    last days. I think your analysis is valid. If the difference between the RV and LV models in the polls are any indication I think we could see PBO exceed expectations nationally and in individual states.

    The loudest cries for war come from those who have never seen one.

    by MadGeorgiaDem on Sat Nov 03, 2012 at 01:02:01 PM PDT

  •  That's more like how he's (perhaps) (1+ / 0-)
    Recommended by:
    exterris

    underestimating Obama's chances than why.

    But I'll tackle the why - as a sports guy, I suspect Nate is just not satisfied until somebody gives (or gets) 110%

  •  I hope we win. I don't want all of this republican (0+ / 0-)

    Meltdown means they are going to come out with big numbers on Election Day!

    "I'm not mad at them (tea party) for being loud, I'm mad at us for being silent for the last two years. Where have we been"? "it was never yes HE can, it was Yes WE can". - Van Jones

    by sillycilla on Sat Nov 03, 2012 at 01:07:01 PM PDT

  •  asdf (1+ / 0-)
    Recommended by:
    nicolemm

    Um - Nate takes ALL of this into account in his model.

    As he has written several times in the past few weeks, as election day approaches, his model will be subject to larger and larger variations as the more time-dependent factors become irrelevant and the emphasis becomes on the most recent polls and trends.

    In other words, by Monday, I would no doubt expect his final percentages to be in the 95% range - given their current position and the trend of recent polling.

    Sadly, everything Communism said about itself was a lie. Even more sadly,, everything Communism said about Capitalism was the truth.

    by GayIthacan on Sat Nov 03, 2012 at 01:07:35 PM PDT

    •  You're right (0+ / 0-)

      I'm not saying that Nate's numbers won't be better by Monday, but they are too low today.

      I'm also not sure that Nate can take GOTV into account in his model.

      I'm also not sure that Nate takes into account the diversity of the states in play.  My feeling from reading his articles is that he accounts for a possible across-the-board bias.  The fact that the 4 big states are so different insulates Obama to a certain extent that I don't think the model will capture

  •  The state polls reflect voters. (0+ / 0-)

    In that regard, if turnout is close to expectations, then I think giving Obama 95% is warranted.  I am cautiously optimistic.  I think the trend lines are continuing to move in Obama's direction.  But turnout is the key.  We're talking about people, and people--like us--don't always perform as the models predict because there are so many variables and unforeseen factors.  For that reason--the same reason that a great baseball team can lose a World Series--we hedge our bets a bit, cross our fingers, and make phone calls to get people out to vote. I'll rest easy when the election is called late Tuesday night or early Wednesday morning.

  •  2 more reasons maybe (0+ / 0-)

    1.he takes all  polls, Wang  takes the median.   Nate's model  more vulnerable to  outliers (like M-D giving R+6 in FL)

    2. he takes demographics into account, i think the majority white pop of most states makes it a (slightly) harder burden for state polls to show O in the lead.

  •  you've described HOW, not WHY. (1+ / 0-)
    Recommended by:
    Matt Z

    ;-)


    "A squid eating dough in a polyethylene bag is fast and bulbous..........got me?" - Don Van Vliet

    by AlyoshaKaramazov on Sat Nov 03, 2012 at 01:45:32 PM PDT

  •  Bigger turnouts reflect Dem victories (2+ / 0-)

    I just returned from Nevada and they expect 80% of eligible voters to vote.  There are more early votes cast this year than 2008.  Yet, the conservative Las Vegas Review Journal paints it a good picture for Mitt Romney!

    But, their political reporter points out the early voters are casting many more for President Obama.  No wonder, the GOP wants to suppress the vote--everywhere!

  •  Nate Silver has always described his model (2+ / 0-)
    Recommended by:
    ericlewis0, Andrew F Cockburn

    As being conservative.  This is a great analysis of how it is.  I personally tend to split the difference between Sam Wang, who assumes that the median state poll isn't biased (for good reason, but it isn't 100% the case) and Nate Silver whose model is skeptical.

    So I'd put it at about 90% which is pretty damn good.

    •  Silver does a great job (1+ / 0-)
      Recommended by:
      nicolemm

      I think his percentages are accurate and they reflect the variability of the electorate.  I don't this election is sewn up yet--Democrats have to match every thrust by the opposition.  They are being outspent big where I live.

      What Silver does so well is get a composite of polls and economic factors and come up with a formula that covers so many factors.  As far as the polls go, commentators have said forever that they are "snapshots" of the electorate.  If you take enough snapshots, the picture is pretty conclusive.  

  •  The polls won't just be wrong in Ohio (0+ / 0-)

    If the polls are systemically biased that bias will show up in all states, because it means that pollsters assumptions about who will turn out to the polls is wrong. The bias might not result in movement of the same magnitude (e.g., if pollsters are assuming that 55% of eligible Latino voters will vote, but only 50% do, it will have a bigger effect in Nevada than in New Hampshire) but the movement should be in the same direction nationwide.

    If the Ohio polls have overestimating Obama's support by 5 points, then it's pretty likely that the polls in Michigan, Wisconsin, and Pennsylvania are too.

    •  You're right about the collection of states (1+ / 0-)
      Recommended by:
      ggrzw

      you're suggesting there.  However, the big 4 states that I'm talking about are very diverse.

      For example, suppose pollsters are systematically overstating Latino turnout.  This would hurt Obama in Colorado, but not in Ohio.  Because of the diversity of the states involved, it would take a whole collection of errors for the polling averages to be wrong in all four states.

      This is certainly possible, but less like than if the states in play were all in the Midwest.

      •  You're right (0+ / 0-)

        The fact that the four key states are pretty dissimilar from each other mitigates the impact of bias, but there are also potential biases that have a pretty broad impact (e.g., misestimating the number of 18-29-year-olds who turn out would have a significant impact on most state polls). I think Silver's model largely accounts for that.

        Don't get me wrong, I'm hoping that the polls are biased in that they underestimate the strength of the Democratic machine in NV and MT, and the Democratic vote share of Latino/as in AZ.

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site