« BACK  |  PRINT

RS

MEMBER DIARY

Election Update

One week to go

Sometimes it’s all about momentum and options, and right now Mitt Romney has both.

It’s a week before Election Day and it seems that the map is expanding for Governor Romney, which means that it’s contracting for President Obama.

In this Election Update I’ll look at the current map and some scenarios that it suggests.  I’ll also give you some ideas about how to interpret the many polls that will come out between now and Election Day.

The Presidential Race

It’s good to have options. 

In some ways, not a lot has really changed in the last week.  Mitt Romney still has a lead in the national polls, and a bigger one among those most likely to vote than among those less likely to vote.  Many of the battleground states are still very close.

But what has changed is that both campaigns have slowly begun redeploying resources away from early battlegrounds such as North Carolina and, to a lesser extent, Florida and Virginia and into states like Wisconsin and Minnesota that were previously thought to be safe for Obama.

This redeployment suggests to us that the internal polls both campaigns are seeing (and we believe these are both more detailed and more accurate than public media polls) are showing that Romney likely to win some of these states that the media still treats as “battlegrounds.”

Let’s look at where the polls are now in a number of states, only analyzing those surveys with interviews conducted entirely in the last week.

In doing this I’m going to look at both the mean of recent polls and the median recent poll.  In some states there is one outlier poll really moving the arithmetic mean.  Looking at the median in those states will give us a better picture of the recent trends.

State

Polling Mean

Median Poll

Colorado

Obama +0.5%

Obama +0.5%

Florida

Romney +1.75%

Romney +2

Iowa

Obama +2%

Obama +2%

Minnesota

Obama +4%

Obama +4%

Nevada

Obama +2.5%

Obama +2.5%

New Hampshire

Obama +0.75%

Tie

North Carolina

Romney +4.67%

Romney +6%

Ohio*

Obama +1%

Obama +1%

Virginia

Tie

Romney +2%

Wisconsin**

Tie

Tie

 *    Because Ohio has so many polls, we were able to use polls only released over the weekend and today.

**  The Wisconsin data is based on just a single poll conducted in the past week. 

Based on these data, there seem to be three courses this election can take.  These are presented in order from what we, today, see as most likely to least likely:

1)   Mitt Romney wins a narrow victory:

  • It seems clear at this point that Romney will win North Carolina unless something significant changes in the next week.
  • Romney is also clearly ahead in Florida and Virginia though not by the same margins as in North Carolina.  If the election were held today, we would predict he wins both.
  • Given that background, Romney would win a narrow victory with any of the following combinations of states:

i.      Ohio and any one other state;

ii.      Wisconsin, Colorado and any one other state;

iii.      Minnesota, Colorado, and any one other state;

iv.      or (least likely) Colorado, Nevada, New Hampshire, and Iowa.

    • Given the trends over the past few weeks and the issues with the Ohio polling that we’ve pointed out previously, we see a narrow Romney victory including a win in Ohio as the most likely scenario at this point.

2)   Obama wins a narrow victory:

  • With the loss of the Southeast, Obama would need to run the table in the Midwestern and Western states in which he currently has narrow leads.
  • While we’re skeptical that Obama can withstand the general trend in this race and that he is even ahead today in Ohio, if Obama “runs the table” in those states (including Ohio), he will win re-election.

3)   Romney wins a landslide:

  •  There is an increasing possibility that Romney could win this race with 300 or more electoral votes.
  • While this is still less likely than a narrow win for one candidate or the other, the fact that some states are moving solidly into the Romney column while others that were previously thought to be solidly for Obama are suddenly becoming competitive is the type of late momentum that leads to landslide outcomes.

So what’s the October surprise? 

With a week to go until the election, there is still time for things to change dramatically.  While the idea of an “October surprise” is often overstated—most late revelations aren’t profound enough to alter the fundamental path of an election—there are two stories that might, just might, change the outcome next Tuesday.

1)   The response to Sandy.

President Obama has cancelled campaign events to return to the White House and “monitor” the course of the storm.  Now, unless Obama is a secret X-Man there’s not really much he can do about the storm and certainly nothing he couldn’t do from Air Force One.   But this does give him the chance to look like a leader and appear Presidential at a critical juncture in the campaign.

It’s possible that, particularly for very late deciding voters, seeing Obama leading the country in a moment of crisis could give him a critical boost.

On the other hand, disaster response is a tricky business especially when major cities that aren’t used to this kind of disaster are involved.  If the response is poor and the federal government looks as incompetent as it often is, Sandy could be the final blow that ends the Obama Presidency.

2)   Benghazi

The evolving story about Benghazi—what the White House knew after the attack and when and, more importantly, what orders were given before and during the attack—has the potential to become a true “October surprise” in this election.

The more information that comes out about the attacks, the worse things look for the Administration.

And it’s pretty clear that groups inside the defense and intelligence community feel wronged by the way the Administration has behaved before, during, and after the attacks and will continue leaking damaging information until Election Day.

The question here is how much coverage this story receives.  So far, many media outlets have all but ignored it—shocking?!  And now we have the real possibility that coverage of Sandy and its aftermath will overwhelm any potential coverage of this story.

If we start seeing extensive coverage of the Benghazi story on CNN and on the network evening news, that too will be the final blow to Obama’s hopes.

Three rules for interpreting polls 

There will be a lot of polls released between now and Election Day.  Here are three rules to help you assess which ones are likely to be the good indicators of how things really stand:

1)   Short field times are better than long.

Some polls, particularly those conducted by colleges and universities, but also some major media outlet polls, have field schedules of four or more days.  Campaign polls will usually poll over two days.  Some robo-polls will poll on a single day.

As we get closer to the election, polls with longer field windows are likely to include data that no longer reflects reality.  In 2008 exit polls, one-in-ten voters said they made their decision in the last week of the election.  This means votes really are shifting day-to-day.  So look for short field-time polls as the best indicator of today’s reality.

2)   Keep an eye on partisanship.

We’ve talked before about polls that have partisan make-up that reflects the 2008 election, or is even more Democratic than that election.

As more and more polls come out, often with divergent results, keep an eye on the partisan make-up of the samples and trust those that are somewhere between the 2004 and 2008 electorates the most.  It’s probably safe at this point to discount polls that exactly mirror 2008 or are even more Democratic than that in their turnout assumptions.

3)   Voter files are better than calling all adults and screening.

This is an old battle in the polling community and one that would be over but for an “old guard” who haven’t kept up with recent research or refuse to put evidence over a theoretical argument about “random samples.”

To simplify the argument, a pollster can either use voters off of the voter file who have turned out in past similar elections as their sample–or they can call random landline (and sometimes cell) phones and ask the person that answers whether they are registered and will vote.

Without going into the arcane details of the argument, let’s look at what happens with the RDD (random digit dial) approach.

  • The Washington Post recently released a poll in Virginia showing Obama up by four.
    • This poll was conducted from 10/22-26 (go back to rule 1 about field windows) and used a RDD sample of landlines and cell phones.
    • 1,504 adults answered the screening questions,
    • 1,309 of them said they are registered voters, and
    • 1,228 said they would vote.
      • That’s 82% of all adults and 94% of registered voters as “likely voters.”

Now, most of our readers already know how ridiculous those numbers are.  But let’s look at the actual numbers:

  • In 2008, Virginia had an adult population of approximately 5.72 million according to the Census Bureau.
    • According to the State Board of Elections, just over 5 million people were registered to vote.
    • According to the Board of Elections, 3,723,260 votes were cast in the 2008 Presidential Election.
    • That’s only 65% of adults and 74% of registered voters.
      • In other words, the Washington Post poll is over-stating likely turnout by 20 points.

Those numbers aren’t really uncommon for polls that use a self-identified likely voter methodology.  So pay attention to how the polls are identifying voters and give more weight to polls that start from the voter file and use previous voting to identify likely voters.

As always, a big thanks to Bryon Allen and the research team at WPA for helping do all the heavy lifting on this analysis.

Please go to http://www.wparesearch.com/blog to subscribe to updates like this directly from us and like us on Facebook at https://www.facebook.com/WPAResearch.

Get Alerts