Sunday, November 04, 2012

48-Hour Warning: Last Call for Prognostication!

In (actually less than) 48 hours now, we will begin voting. Technically, many of us already have, in a pattern that, incidentally, seems to favor Obama so far. But I digress. My point is, the game is almost over. Below are the leading indicators as they currently stand.

Popular Vote

RCP polling average (a straight average of selected recent polls):

HuffPost Pollster polling average (an average of a wider set of polls, with adjustments for state-national variance and house bias):

TPM polling average (selection of polls, and with regression trends added in):

If you take the average of all three, it looks like this (also showing the collective movement for both Obama and Romney over the last 10 days):
So, in sum, slight advantage Obama, but so small compared to the margin of error that it could just as easily be either candidate ahead or behind, probably within 1 point of each other. 

Now, there are those who disagree. Throughout the campaign, the average of state totals have seemed to add up to a different number than national polls have been showing. Nate Silver at FiveThirtyEight has a good argument that the state totals are the more reliable of the two, and indicate a stronger lead for Obama than national polling averages are showing. Simon Jackman, a Political Science professor at Stanford, concurs, using a similar analysis. Based on this line of thought (and the math behind it) Silver's model indicates the following result for election day:

  
Electoral Vote

Even as its polling average shows a slight lead for Obama, the state averages at RealClearPolitics show Obama with leads in 9 of 12 ostensible swing states. His lead is less than 2% in only two of these (Colorado and New Hampshire). RCP's "no toss up" map currently shows:

Over at HuffPost Pollster, their model, which has been updated with 120 swing state polls in the past week, currently shows:   
To paraphrase the math, Obama is overwhelmingly likely to reach at least 277, with a strong chance of up to 303.

FiveThirtyEight similarly updates its electoral model with all new polls, and then makes adjustments for house bias and historical dynamics in each state. In the past two days, they've only found a Romney lead in 1 out of 20 swing state polls released Friday, and 2 out of 21 released Saturday. Their map currently looks like this:
  
The only states on the map that have below an 80% chance of an Obama victory are Colorado, New Hampshire and Virginia, and the lowest (Colorado) still shows a 68% chance. On this basis, Silver's model is predicting an overall 85% chance of Obama victory, and a map that adds up to 303-235.  

Finally, we have 270towin, which has built up a database of all recent state polling, and uses it to run 10,000 simulations a day. Their most recent batch comes out like this:

Based on all available evidence, Obama seems very likely, whatever the national popular vote, to win the electoral college.

Oddsmakers

Of course, it isn't all poll averages and simulation models. There are several betting markets that aggregate the bets of large numbers of participants (like, really large- over 100,000 on an average day). Two of the largest are Intrade, which currently shows a 66% chance of Obama's reelection:

And Betfair, who's current odds are:

These are odds against, so if you translate them back to percentage terms, it comes out to around 76%-24%.

The betters are predicting very high odds for Obama winning. 

Approval Rating

One last data point. RCP also tracks Obama's job approval rating over time based on an average of multiple surveys, and it currently adds up to:

Obama is just a whisker under 50% approval, and his approval leads disapproval by nearly 3 points. This is much more likely to signal reelection than not. It also, perhaps not coincidentally, is very similar to where Bush was on election day in 2004 when he won a narrow, but hardly ambiguous, victory:


Final Thoughts

Many people are presenting the race in this final stretch as "too close to call". This certainly makes sense for reasons of better news leads and partisan positioning, but I don't believe that it's accurate. As I shared in my "one week to go" posting, I learned a hard lesson in 2004: if the preponderance of numbers are pointing in a direction, believe that direction over any story you may see, no matter how compelling a case it makes,  about why the preponderance of numbers is actually wrong.

With less than 48 hours to go, the preponderance of numbers indicate that Barack Obama will be reelected. 

Since I've been throwing everyone else's forecast around here, I'll add mine to get some numbers where my mouth is. On the popular vote front, taking the average of RCP, Pollster and TPM as my starting point, and assuming a 1% 3rd party vote (which is pretty typical of years that don't have an Anderson, Perot or Nader 2000), we get 3.8% still undecided. To the extent that there is any evidence of closing momentum, it seems to be slightly toward Obama. So I tilt the split of the undecideds very slightly in his favor. And on this basis, I'm going to predict a popular vote margin of Obama 49.7% to Romney 49.3%.


In the electoral college, I'm going to go with all the states where Obama currently leads by more than 1% going to him, all the ones that Romney leads by any amount going to him, and Colorado, where Obama leads by less than 1%, and the only state where the early voting trends do not seem to favor him, going to Romney. On that basis, I predict Obama 281 to Romney 257.   

As anyone who's worked with me in my finance jobs over the past 15 years (eeek-is it really that long?) will tell you, I tend toward a (numerically) conservative bias in my forecasting. So, to the extent that there is upside to this forecast, I would expect it to be in Obama's favor.

And that is my final "before" word on the subject. My next stop will be live blogging election night. Join me then!





No comments: