Friday, November 2, 2012

The Big Ohio Update

I'm going to cover a lot of ground in this post, so hang in there.

Polling update

One new poll comes into the RCP, this time from University of Cincinnati showing Obama up by 2. Let me warn you up front, this is an outlier.  The Independent preference is not following any of the other polls, and I am very suspicious of it.  The D/R/I of the poll is 42.68/42.07/10.87 with Independents favoring Obama by 14 points.  Since they only polled 140 Independents, I am really suspicious of the results.  When I put this poll into the models, they look very very bad for Romney.

O+2 - Current result
R+1.25% - 2004 turnout model
O+7.18% - 2008 turnout model
O+1.68% - 2010 turnout model
O+2.15% - 2012 registration model
O+4.32% - D+3 turnout model

The Independent preference is simply wrong.  I really just can't put any faith in the poll at all.  Trying to figure out what could be causing these results, I noticed that 40% of the poll respondents are from Northeast Ohio.  According to the numbers I can find, this section of Ohio only represents 30% of the population, but it is the stronghold for Obama.

Despite how unbelievable this poll is, I will go ahead and include it in the model and look at the results:

Polls included:
O+3 - Survey USA
R+2 - Rasmussen
O+4 - PPP
O+5 -Quinnipiac
O+2 - Purple Strategies
O+4 - CNN/ORG
O+2 - ARG
O+1 - Gravis
O+2 - U of Cin.

O+2.33% - Current RCP Average
R+6.76% - Average using the 2004 turnout model
O+2.72% - Average using the 2008 turnout model
R+3.53% - Average using the 2010 turnout model
R+2.84% - Average using the 2012 registration model
R+0.63% - Average using the D+3 turnout model

This one poll drops Romney's average across all the models by about .7%.  Despite this outlier, Romney still leads in Ohio by better than 3 points if we have an Even turnout, and is tied with a D+3 turnout.  This is really the crux of the election.  Will the Democrats have a +3 turnout edge or not.

Credible Poll Averages

In 2004 I was hanging out on a blog that was working through all the polling.  Things were a lot like this weekend before the election.  New polls were cropping up all the time showing Kerry surging.  We had a polling professional on the site who would look into the polls and declare them credible or junk.  He would simply ignore junk polls and only pay attention to the credible ones.

I'm going to do something similar.  If I look at the internals and history of these polls, I can see which polls are getting credible data, and which ones are outliers.  Even polls that are D+8 can still be credible, if the internals of the poll are consistent with the trends we are seeing.  In these cases they are simply sampling too many Democrats.

Using this screen, I'm keeping the following polls and running the models:

O+3 - Survey USA
R+2 - Rasmussen
O+4 - PPP
O+5 -Quinnipiac
O+2 - Purple Strategies
O+1 - Gravis


In the end, I am throwing out the CNN and U of Cincinnati polls because their Independent preference results don't match the rest.  I am also throwing out ARG because their Ind +21 result for Romney is also not credible.  The remaining polls range from Romney +2 to +11.

O+2.17% - Current RCP Average
R+7.78% - Average using the 2004 turnout model
O+1.89% - Average using the 2008 turnout model
R+4.50% - Average using the 2010 turnout model
R+3.78% - Average using the 2012 registration model
R+1.55% - Average using the D+3 turnout model

These results do a better job of reflecting the current state of the race.  If we have a D+3 turnout, then Romney will win Ohio by a close number.  If we have an Even turnout, Ohio will be called early with Romney winning by about 250,000 votes.

Early voting

Below are the reported numbers of early votes cast in Ohio as of November 1st.  This does not represent all of the counties, apparently those are not online.  Overall the counties listed in this picture represent 79% of all of the votes cast in 2008.


I spent a few hours with these number this morning, and can draw a few conclusions.  The major Obama strategy is to use early voting to run up the vote totals.  I have found a quote by a Professor at University of Dayton that McCain actually won the vote on election day.  Where Obama developed his 4.6% lead was in early voting.  I tried to find evidence this morning that confirmed this statement, but couldn't (just a lot of people quoting him).

Assuming he is right, then the 260,000 votes that Obama won by in 2008 came exclusively from early votes.  Now in 2012 the number of voters statewide has dropped 5.9%.  Unfortunately for Obama, Cuyahoga county dropped 19%, by far the largest loss of any county in the state.  But assuming a statewide distribution, Obama would need to bank about 245,000 extra votes in early voting to replicate his 2008 win.

So if we look at actual votes received by county and compare them to the early vote percentage in that county, we find that 1,063,851 votes have been received so far.  But to match the 2008 numbers we would expect 1,423,425 early votes to be cast.  Early voting is currently down 359,574 votes in these counties.  Remember, these counties only represent 79% of the 2008 electorate.

Using the vote share that Obama enjoyed in each of these counties, we can estimate how many early votes Obama received in each county.  This estimate favors Obama slightly, since his actually numbers were higher due to his focusing on this strategy.  This reveals Obama's under performance in early voting, how many votes he should be getting in each county, and how many votes he still needs to meet his turnout goals.

Adding it all up, as of November 1st, Obama is 208,598 votes under his targets in these counties.  extrapolating to account for all counties, he is 261,955 votes under his goals.  We can drop that number a little to account for population loss and say he is 246,499 votes short of his target.

In other words, his early vote performance advantage is wiped out, and he is tied.  This is before considering any other factors.

Let's take a closer look at his strongholds.  These are the large urban areas that he won convincingly in 2008.

  • Cuyahoga County voted for him 70% and had 265,000 early votes - currently 69,000 votes short
  • Franklin County voted for him 61% and had 250,000 early votes - 73,000 votes short
  • Lucas County voted for him 67% and had 63,000 early votes - 20,000 votes short
  • Mahoning County voted for him 66% and had 40,000 early votes - 9,000 votes short
  • Summit County voted for him 59% and had 60,000 early votes - 27,000 votes short

Compare this to one of the McCain counties

  • Warren County voted for him 32% and had 24,000 early votes - 2,000 votes short
A lot of McCain counties aren't even reported, like Butler County that had a total of 173,000 votes and voted for McCain by 60%.

Independent votes

In 2008, Independents made up 30% of the electorate in Ohio and favored Obama by 8 points.  This is a vote total of:

  • Obama - 923,000
  • McCain - 787,000
  • Advantage Obama -  136,000 votes
This year, that trend is reversed.  Romney is holding an average advantage of 8 points with Independents.  Assuming a 5.9% reduction in overall votes but the same 30% share of the electorate, this becomes:

  • Obama - 740,000
  • Romney - 869,000
  • Advantage Romney - 129,000 votes

 

Summing it all up

The early voting results suggest that Obama and Romney will enter Election Day tied in Ohio.  In 2008 McCain under performed what the GOP was able to produce in 2010 by 6 points.  This represents a total of about 321,000 votes (6% of 2008 - 5.9% loss in population) that Romney should be able to produce through enthusiasm.

Let's assume that the Democrats are also able to increase their election day totals accounting for new voters (scoff) and fraud.  I will go ahead and spot them 10% of their early vote totals in their strongholds to account for double voting on election day.  This would give them an additional 50,000 votes.  Also giving them a 3% increase in election day turnout (3% of 2008 - 5.9% loss in population) I will give them another 160,000 votes.

Romney turnout increase = 321,000 votes
Independent advantage = 129,000 votes
Obama early vote advantage = 0 votes
Obama turnout increase = 160,000 votes
Obama fraud = 50,000 votes

Romney will win Ohio by 240,000 votes.

17 comments:

  1. From your keyboard to God's eyes, Dave. I hope it's just my normal pessimism kicking in, but I've got a bad feeling about this thing. Thanks for the update, as always.

    ReplyDelete
    Replies
    1. It's your pessimism. This is a repeat of 2004, where all the media hype and spin was showing Kerry surging on the last weekend.

      It comes down to turnout, end of story.

      If the Democrats can get a D+3 turnout, then they have a chance of a very close election. But I don't see how they can get there. I'm willing to hear arguments, but all the evidence says they can't.

      Delete
    2. I just keep thinking about those bloody electronic voting machines registering Romney votes as votes for Obama. And wonder how many people have *not* noticed it when it happened.

      Delete
    3. Notice I gave Obama 50,000 extra votes to account for that, and he still loses?

      Delete
    4. If the Democrats can get a D+3 turnout, then they have a chance of a very close election. But I don't see how they can get there. I'm willing to hear arguments, but all the evidence says they can't.

      You don't think D+3 is realistic given 2008's DRI, which was a composite of roughly 4-5% Republican under-performance and 3-4% Democratic growth?

      Delete
    5. No I don't, the Dems would need to over perform 2010 where they performed quite well. They are under performing in early votes, while GOP enthusiasm is at or above 2010 levels.

      An Even turnout is much more likely, and I can't see them getting to D+3 when they are tied in early voting. EVs are where they are supposed to drive up their numbers.

      Delete
  2. I have come to realize that the polling situation in 2012 is a lot like the global warming debate. The models would desperately like us to believe that the medieval voting period (2010) never, in fact, happened.

    ReplyDelete
    Replies
    1. LOL...that's good...the Medieval Voting Period. If it ever really happened, Wikipedia would try to erase it from the records. :-)

      Delete
    2. Feel free to spread it around and credit me!
      I'm a stay-at-home dad (which means an unemployed smart-ass) and I feel like it's one of the more clever things I have come up with lately.

      Delete
    3. I should also note I came to this conclusion with the help of Michael Barone, who has shown a very close correlation between off year "popular vote" for House elections and popular vote in presidential elections:

      http://american.com/archive/2011/july/what-the-popular-vote-for-the-house-tells-us/

      Delete
  3. Sounds about right, I'll go with 300,000. I think the R turnout will be colossal.

    ReplyDelete
    Replies
    1. I do too. The irony is this post is my pessimistic take on it.

      Delete
  4. With your Ohio models, do you mean D+3 in Ohio or D+3 nationally??

    ReplyDelete
    Replies
    1. D+3 in Ohio. 2010 was R+1 in Ohio.

      Delete
    2. You think it'll be that on Tuesday or perhaps even?

      Delete
    3. I think it will be even. I'm hoping it is R+2.

      Delete
    4. I don't know if we'll be R+1 in Ohio, but that 20-30K at the Romney rally there last night sure makes me think we might be. :-)

      Delete