Monday, November 5, 2012

Final Projections & My Guide to the Election

Final Projections for President
8 Hours Until the Election-Day Polls Open
Final Projected Popular Vote Total: Obama +0.9%
Final Projected Electoral Vote Total: Obama 303, Romney 235
Election-Eve Betting Odds: Obama 67%, Romney 33%
Election-Eve Popular Vote Betting Odds: Obama 60%, Romney 38%, within 0.5% - 2%

Of note on the election eve:
(1) No earth-shattering changes on election eve in the state-by-state count.  The closest and hardest-to-call statements (within my historic margin of error) are Virginia, Colorado and Florida, in that order.  Neither of those 3 are likely to be the tipping point state in the election.
(2) Romney HAS made real gains in Pennsylvania with his late campaign effort and spending, but appears to be falling just short.  Interestingly, in my calculation Pennsylvania IS, at this point the tipping point state in the election.  One to watch on election night.
(3) Minnesota is effectively off the table as a swing state
(4) The popular vote vs. electoral vote polling disparity has disappeared.  My state-by-state projections, when run through the 2008 electorate, produces a margin of 1.0% for President Obama.  The national polling implies a 0.9% advantage for him, insignificantly different.

I remain comfortable with my projection that the President will carry the day.  But this is certainly not 2008, where I make that projection with near 100% certainty.

One note that I should make is that I am making all of my final calls on election eve.  Polls will continue to be released during the day tomorrow and many websites "final" calls are based on these polls.  In my mind, if you don't project it ahead of time, it isn't really a projection.  So, the other websites in many cases have an advantage of newer information, but I will continue to benchmark my performance against what they call their "final" projections.

It will be extremely difficult to meet the accuracy of my popular vote projection in 2008, when I was dead-on to the final result (although I ironically believed that I had missed by 0.7% based on the preliminary election results.)  I don't have any aspiration of being so exact again.

Last time I called 48 states right.  I'd like to match that benchmark and believe I could, although I find 3 states very difficult to call, so a reasonable range would be to get 47-50 states right.

Of course, either candidate could significantly outperform the polls and I could be far worse than last time - we'll have to tune in to see.

Final Projections for the Senate
Projected Senate composition: 52 Democrats, 46 Republicans, 2 Independents (effective Democratic working majority of 54-46)
The Democrats are poised to retain the Senate.  The closest races, in Virginia, Wisconsin and Montana will provide the margin.  The Republicans best chance to pick up the Senate would be to win in the states that I project plus win in very-close Wisconsin and Virgina, pick up the tough fight in Massachusetts and find one other state to take, plus win the Presidency, a near-impossible task unless there is an overwhelming surge for Romney, with coattails.

Final Projections for the House
Final generic ballot average: Republicans +0.1%
Projected House composition: 239 Republicans, 196 Democrats

Republicans will comfortably retain the House, thanks to very favorable redistricting and a strong incumbency position.

Other Projections
Here is what others are projecting in the electoral college:
Fivethirtyeight (Nate Silver) - Obama 332, Romney 206
realclearpolitics (no toss-ups) - Obama 303, Romney 235 - Obama 281, Romney 206, 51 Tied - Obama 290, Romney 248
Fox News - Obama 202, Romney 192, 146 Toss-Up
CNN - Obama 237, Romney 206, 95 Toss-Up
PBS - Obama 247, Romney 246, 85 Toss-Up
Karl Rove - Obama 184, Romney 180, 174 Toss-Up (seriously?  174 toss-ups?  I could've made that projection 2 years ago!)

Of the sites that endeavor to make a projection in every state, there only real area of disagreement is about the closest state - Virginia.  Virginia is certainly extremely close and very hard to call, so this is understandable.  The other states we are in violent agreement, mainly because we all read the same polls.  So we are only going to be wrong when the polls are in agreement if the polls are systemically wrong.

Poll Closing Times (all times in Eastern Time)
Of critical importance as you are following the coverage is when polls close in each state as this will be the first look we get at exit poll data in those states as well as the ballots actually beginning to be counted.  Here are time horizons for all 50 states.  Note that I have sorted them based on when the LAST polls close - as an example the majority of polls in Florida close at 7 PM ET, but the Central Time Zone polls in the panhandle close at 8 PM ET.  Typically, but not universally, states will not be called by the networks until all polls are closed.

7 PM Closing Time (6 states including 1 battleground)
Indiana (Eastern time zone polls close at 6 PM)
Kentucky (Eastern time zone polls close at 6 PM)
South Carolina

7:30 PM Closing Time (3 states including 1 battleground)
North Carolina
West Virginia

8 PM Closing Time (17 states/territories including 3 battlegrounds)
District of Columbia
Florida (Eastern time zone polls close at 7 PM)
New Hampshire
New Jersey
Rhode Island

8:30 PM Closing Time (1 state)

9 PM Closing Time (12 states including 2 battlegrounds)
Michigan (Eastern time zone polls close at 8 PM)
New Mexico
New York
South Dakota (Central time zone polls close at 8 PM)
Texas (Central time zone polls close at 8 PM)

10 PM Closing Time (5 states including 1 battleground)
Kansas (Central time zone polls close at 9 PM)

11 PM Closing Time (6 states)
Idaho (Mountain time zone polls close at 10 PM)
Oregon (Mountain time zone polls close at 10 PM)
North Dakota (Central time zone polls close at 10 PM)

12 AM Closing Time (1 state)
Alaska (Alaska Standard Time polls close at 11 PM)

As you can see, Virginia and Ohio close pretty early in the night, so there are a lot of clues that we can glean from the early returns there.  If Mitt Romney is strong there, it could be a good night for him, or at least a long night and a close race.  If Obama can win in either of those 2 states, it is probably over.  Pennsylvania comes half an hour after Ohio and as my current projected "tipping point" state, will also be important to watch.

If we are still waiting around for the results of our last battleground, Nevada, which closes at 10 PM, we know we have a nail-biter election.

Things to Watch For:
(1) Turnout, Turnout, Turnout
The key debate among pollsters and the key talking point from the GOP for plausible path to victory for Mitt Romney has been that pollster are significantly overestimating youth and minority turnout, both of which were very strong for him in 2008.  If the electorate looks like 2008, it will be a good night for Obama.  If it looks like 2004, it could be a good night for Romney.

(2) Don't Just Watch the Totals, Look at WHERE the Votes are Coming From
We can get ourselves in trouble in early returns if we simply take the partial totals at face value.  To understand the impact on a projection, we need to look at the returns in the swing states by county and precinct and overlay them with the 2008 results, then understand what Obama's margin was in 2008 in that state for comparison.

For instance, in Ohio, Barack Obama won by 4.6% in 2012.  In Hamilton County, he won by 6.9%.  Therefore, if we see early returns in the state from Hamilton County that show him up by 3%, it probably implies a very close race.  If we see him trailing in Hamilton County, he's probably in trouble.

This concept is very important in Pennsylvania, which is essentially composed of Philadelphia (extremely Democratic), the suburbs and exurbs of Philadelphia (swing areas) and the rest of the date (strongly Republican.)  You could see an early lead that shows one candidate 30 points ahead and it still might not mean that he carries the state.

(3) Take the Exit Polls with a Grain of Salt
There has been a systemic problem for several election cycles of exit polling being biased towards the Democrats.  Early exit polls made many believe that John Kerry was going to soundly beat George W. Bush in 2004.  Similarly, in 2008, the raw exit polling data showed a much more massive victory for President Obama than actually materialized.

Unlike a pre-election poll, exit polls are very difficult to correct for participation bias - certain people are more likely to talk to pollsters coming out of a booth than others.  Also, it is much harder to be scientific  in the moment, without time to assess and reweigh demographics for this effect.

So if you see early indications of exit polls that show President Obama much stronger than the pre-election polls, be wary that they mean anything.

(4) Is There a Surprise State?
Could a state no one is thinking is in play and is therefore lightly polled actually show up as a battleground?  Since there isn't much polling in states that we all believe are safe in one direction, these surprises can happen, although they are rare.

Could Mitt Romney shock everyone and win a Maine or a New Mexico?  Could Barack Obama pull off a shocker in Georgia or Montana?  They seem unlikely, but stranger things have happened.

(5) Watch Intrade
The state-by-state and national markets are very liquid during an election, quickly incorporating all the new data coming in from multiple sources.  This is not to say that they are always right, but they are a great barometer of who is in the hunt.

Betters will make a "call" with their money long before gun-shy (since 2000) networks are willing to make a call in a state.

All the debates, campaigns and ads are done (well, maybe not all the ads, but almost!)

Nothing left to do but:
(1) Vote
(2) Watch the results roll in
(3) Congratulate the winner

I hope you will join me tomorrow night for live-blogging during the returns.  Please get out and vote for your candidate.  And if you like this site, tell your friends.

No comments: