A month ago, we brought you the battleground states based on the site's model, and it's time for an update. Recall that the projection is not what would happen if the election were held today (that's the snapshot), but what 538 projects will be the November result.
Based on the data, we project 11 states inside of five points:
Five points is somewhat arbitrary (why not four, six, seven, etc.). One argument supporting five points is that ground organization and GOTV tends to not be able to make up much more margin against an accurate poll of voter preference. Whatever the dividing line, it is important to pick one number and stay consistent over updates. We've used five before; we'll stick with it.
Since the mid-June update, New Hampshire and North Dakota have moved outside the five-point range, toward Obama and McCain respectively. They remain the next two closest states outside the five-point dividing line (Pennsylvania ties New Hampshire at a 6.0% projected Obama win). Colorado has moved into the range, and all the other states have remained in place. Within the battleground group the states have jostled for ranking position, but given that Nate's tweaked the model a bit in the interim it's not worth attaching much importance to the tiny changes.
As for the Penumbra States, the model projects, eight McCain states will finish between 5-10 percentage points, for 81 EVs:
Seven Obama Penumbra States project between 5-10 percentage points, for 67 EVs:
Obama's safe base states (projected double-digit wins) total 175 EVs, while McCain's safe base states total 79 EVs. We project Minnesota as an 11.6% Obama win.
Adding in the 5-10 point projected wins, the totals are Obama 242, McCain 160. Last month, this projection was Obama 247, McCain 157.
Obama needs 28 EVs out of the battleground group for a clean win, 27 EVs for a tie and a messy but near-certain win. We predict Obama will win 51 EVs out of the 11-state battleground group.
The 538 mid-July projection: Obama 293, McCain 245.
Other 538 breakdowns:
Polling Average: Obama 309, McCain 229
Trend-Adjusted: Obama 309, McCain 229
538 Regression: Obama 298, McCain 240
Snapshot (if the election were held today): Obama 303, McCain 235
Wednesday, July 16, 2008
538 Battlegrounds as of Mid-July
-- Sean Quinn at 4:16 PM
Labels: battleground states, electoral math
133 comments
I think the changing demographics plus the Mark Warner landslide will push Virginia into the Obama column. Just my take.
Respect for the model, I guess, but it seems a bit nonsensical to move ND out of the battleground classification when the only change is a moderation in the Obama trendline adjustment nationally, considering that the only ND poll to come out in a long long time came out both recently and strikingly pro Obama.
Not that it matters, of course. Thanks for the update =).
You are using the term WIN% differently than Nate does. What you are calling WIN% is what Nate calls Margin.
This may cause confusion.
I doubt the potential Mark Warner landslide will do much, Charles, unless there are personal fans of Mark Warner that might not have otherwise voted, who will be motivated to turn out to vote for him for Senate. I don't see that happening... if you're a big enough fan of a politician to vote for Senator, you're already going to want to vote for President.
What might hurt McCain more in Virginia, I would think, is the same motivated/unmotivated numbers we see elsewhere. But that's independent of Mark Warner fans.
I wonder how, and could you enlighten us, how are Nader and barr figured in here? Pollster shows them pulling Mccain out of any type of contention in most every state they appear. And nationally - they pull McCain 10 points behind Obama.
John: I think what Sean means by Win % is actually "projected margin of victory", which corresponds to the
"Projection" lines in the table on the right.
alex: ND is still fundamentally underpolled. (Though I'm grateful for what we got... how long did we have to stare at that lonely, shrinking February result?).
at Sean: It would be cool to see what the battleground states were a month ago, using today's methodology. That would be the real way to look for changes.
Interesting to observe the movement. I confess to being confused by the fact that North Dakota has moved out of the five point win percentage 'battleground' window towards McCain, considering that the only news out of the state has been a surprisingly favorable poll for Obama showing a dead-heat (as well as similarly auspicious results in neighboring states of Montana and South Dakota.
Holy Smokes! NYT article says Obama will hold a press conference tomorrow at 10 ET. Supposedly, he will be discussing his dual faith in Christianity in Islam. It's going to hit the fan now!
Field has Obama 54 - McCain 30 in CA. Their previous(May) poll was +17 for Obama.
No way. This will confirm a lot of people's suspicion that he does have strong ties to the Muslim community. Look for polling numbers to plunge but it should remain close!
I enjoyed this very much. Thanks Sean.
Barack needs to start advertising in South Dakota. It has to be cheap. Plus although they have different media markets, advertising in South Dakota would help in North Dakota. Remember that Dakotans divide their own states east and west (mentally). "Only" the remaining 48 states divide them north and south.
Nice summary, Sean. Thanks for laying it all out so clearly.
I do have one minor objection, though: at a winning projection of 0.1%, VA would fall into recount territory.
Perhaps it would be a good idea, even for these summaries, to separate out states that are "too close" to call. Pick a close (if arbitrary) number such as 1% or 0.5%, and if a state falls within that range, then mark it as tied or potential recount for the projection summary. That would leave the current 538 EV projection as:
Obama 293, McCain 232, Too Close To Call 13.
Nate uses <= 0.5% as the Recount marker in the Scenario Analysis, so that might be the best number to use for separating out states that are essentially tied, for consistency -- if you decide this argument has any merit.
.
Holy smokes! McCain will hold a press conference tomorrow where he will be discussing how his supporters have their heads thoroughly shoved up their asses.
It shouldn't affect him at all in the polls because, well, his supporters have their heads thoroughly shoved up their asses.
What's the difference between trend-adjusted and snapshot?
Looking at this it is difficult to justify putting in much of an effort in Florida.
Michigan is probably a lot safer than indicated. There is probably still some residue resentment over skipping the primary, and Hillary did a good job of blaming Obama. That will fade.
That leaves only 10/11 EV's. There has to be safer & cheaper options than fighting for the 27 of Florida.
Besides there has to be poetic justice. The local Democratic Party did absolutly nothing to stop Hillary taring Obama with their mess.
Let them be ignored for the General Election too. No coat tails, no advertising revenue.
Tybalt: I doubt the potential Mark Warner landslide will do much, Charles, unless there are personal fans of Mark Warner that might not have otherwise voted, who will be motivated to turn out to vote for him for Senate.
No, Charles has a point. It's not that personal fans of Warner will push Obama over the winning line in VA - as you note, they probably would have come out to vote for the presidency anyway.
Instead, Warner might have some coattails among independents in the voting booth, and may help in locally getting out the vote for Obama in a way that a less popular Senate candidate might not.
One could argue that such effects will be balanced out by ticket splitters, and that is a possibility, but given the general anti-Bush and anti-Republican trend these days, the former scenario seems more likely.
.
1magine has a point. Pollster.com shows a 6.8% combined vote for Nader and Barr in a four party race with McCain losing the most both in nationwide and in state polling. The discussion of the decline in third-party support as elections approach has be rather thin on this and other sites. I would like to see more evidence that none of the "other" candidates will have any impact on the election this time around.
Why is the regression added to the trend-adjusted polling average *before* applying the discount to Obama's lead for the difference between Snapshot and November Projection? If the "Regression" is based on relatively stable demographic factors, why apply the time-discount to it instead of taking it as a baseline that will not change from here to November, unlike the more volatile polls? Wouldn't it make more sense to average the Regression with a poll-based and time-discounted projection for November?