Ohio as Icing on Obama’s Victory Cake
No, I’m not calling the 2008 Presidential election now for Senator Barack Obama, my candidate of choice. There’s too much time remaining. Too many things can happen. (And I’m not confident there will even be an election.)
What I’ll do here is point you to Sean’s June 16 post on the FiveThirtyEight.com electoral projections blog titled “538’s Battlegrounds as of Mid-June“, provide a little commentary, and do one of my favorite things — play with numbers.
Here is the short version: Assuming Obama wins all of the states in which he presently leads in the polls, except for Ohio, he would have 297 electoral votes (EVs), 27 more than is needed to win the election.
OK, if you’re game for some arithmetic, stick around. As an enticement I’ll eventually point you to FiveThirtyEight.com’s projection of the outcome of the election in terms of percentages and explain why the current numbers are better than it sounds for “The Big O” and are likely to improve barring unforseen events.
But first some background on FiveThirtyEight.com: Nate Silver’s site rose to prominence in the progressive political blogosphere during the primary season with a series of uncannily accurate projections. Posting on Daily Kos as Poblano, Silver attracted quite a following. Silver is not a pollster. Rather, he has developed a statistical model for electoral projections based on poll results. As Richard explains in the comments on Sean’s post:
Nate’s statistical model … is not based purely on … current performance in polls at all. Nate takes into account historical trends and demographics in each state through his regression analysis, and he compensates for historical trends in polling error by running daily simulations.
What we do over there [at Baseball Prospectus] and what I’m doing over here [at FiveThirtyEight] are really quite similar. Both baseball and politics are data-driven industries. But a lot of the time, that data might be used badly. In baseball, that may mean looking at a statistic like batting average when things like on-base percentage and slugging percentage are far more correlated with winning ballgames. In politics, that might mean cherry-picking a certain polling result or weaving together a narrative that isn’t supported by the demographic evidence.
538, by the way, is the total number of EVs up for grabs.
Now back to Sean’s post. He starts with the premise that five percentage points is the most a campaign’s efforts in the field can make up against an accurate poll of voter preference. He then identifies 12 battleground states in which FiveThirtyEight’s regression analysis of polling data shows the race to be within five percent either way:
- Missouri — Obama by 0.3%
- Nevada — McCain by 0.4%
- Indiana — McCain by 0.7%
- Michigan — Obama by 1.6%
- Virginia — Obama by 1.8.%
- Florida — McCain by 2.0%
- New Mexico — Obama by 2.5%
- New Hampshire — Obama by 2.6%
- North Carolina — McCain by 2.7%
- Ohio — Obama by 2.9%
- North Dakota — McCain by 3.2%
- Montana — McCain by 4.2%
That leaves 38 states plus DC with margins of five percent or more. Of those, the states in which Obama is ahead total 247 electoral votes (EVs) and the states in which Senator John McCain is ahead total 157 EVs. 270 EVs are needed to win.
So to win Obama would need an additional 23 EVs from the battleground states. Where would they come from? I think Michigan (17 EVs) and New Hampshire (4) are gimmes. FiveThirtyEight gives Obama a 63% chance of winning Michigan and a 73% chance of winning New Hampshire. I know those don’t sound like gimmes, but they really are pretty good for Obama. I’ll explain in a bit.
Michigan and New Hampshire would up Obama’s total to 268 EVs, just two shy of the magic 270. Of the remaining battleground states, I really like his chances in Missouri (11 EVs), Nevada (5), and New Mexico (5). Any one of the three would do. For these states FiveThirtyEight’s projections are comforting: Obama’s chances of winning are 61%, 56%, and 69%, respectively. Way back when I would have been able to calculate the chance, based on these projections, that Obama would win at least one. I’ve long since forgotten how. But I betcha it’s pretty high.
So there you have it: A scenario for Obama to win without winning Ohio and for that matter, Florida, the states that cost the Democratic Party control of the White House in 2004 and 2000, respectively. Wouldn’t it be excellent to not have to pin Obama’s hopes on those infamous states. Except, you know what? I predict Obama will win Ohio. FiveThirtyEight likes his chances in Ohio too — 82%. That would be a nice gob of icing on Obama’s victory cake!
Alright, I promised to reveal FiveThirtyEight’s projection, on a percentage basis, of the outcome of the election. They currently project the chance of Obama winning is 74.3%. Pretty good, right? But if you’re like me you won’t rest easier until it’s up above 90%.
Well here’s the thing — barring unforseen events, the projected chance of Obama winning is likely to increase as the election draws nearer. Why? Take it away again, Richard:
Nate’s projections are based on 10,000 daily simulation runs using current polling data and an error figure based on analysis of historical accuracy of polls, which Nate described most recently in his post (We don’t know as much as we think (Big Change #1). Predictably, this error varies with time, becoming smaller as the election nears.
While this explanation should make Obama partisans feel better, keep in mind FiveThirtyEight’s projections are based on a statistical model and do not — cannot, really — take into account the effect of possible future scandals or even polling momentum. As I said at the beginning of the post, too many things can happen between now and November 4. If you’re an Obama supporter, this is not the time to coast. I encourage you to support his campaign by donating your time or money.
As of this juncture, though — four and a half months before the election — Obama is establishing himself as the clear front-runner. Just two days ago when I started writing this post, FiveThirtyEight was projecting a 65% chance of Obama winning. Now the jump from 65% to 74.3% isn’t due to the shrinking of the error figure over time Richard described. Two days isn’t much time after all. Rather, the increase is due to unexpectedly good (for Obama) Quinnipiac poll results in Florida, Ohio, and Pennsylvania. 🙂