All the main election polling sites (the ones I frequent) now project Obama to win the Presidency, though there is disagreement over which states remain in play. But McCain's camp is claiming they still have a chance, scaring Rachel Maddow and other Cassandras like me.
Remembering that Kerry led in the polls this time last year, we might want to view them with some more skeptical assumptions—a sort of worst-case scenario for Obama—and see where the battlegrounds really are, using two different "safety" measures and data from several sites, which aggregate the polls in different ways.
The data show McCain can win, even without an "upset" state like PA. However, he has far more hurdles than Obama. A McCain win would require a massive opinion swing, as he starts from a much lower base (more than 100 electoral votes down, however you count, and he'd have to overtake Obama in "battleground" states that many polling pundits see as Obama-safe. Nevertheless, only one site has Obama with the 270 EV he needs after you toss out the leaners. And the lean may mean less than you think. In 2004, 11 states changed 7 points or more in the polls over the week prior to the vote. Let's look at the numbers.
Let's see if I can talk myself down, always bearing in mind that everyone may be wrong.
In evaluating pollsters' past performance, Nate Silver cautions:
Notwithstanding that one result isn't anywhere near enough information to conclude that a pollster is strong, the 2004 election was perhaps the easiest one in history to forecast. The electorate was highly partisanized with few undecideds, and both bases turned out in roughly equal numbers; it wasn't just IBD/TIPP that got it right -- almost everyone did. This election is considerably more difficult to poll, and it's exposing the weaker pollsters.
50% Scenario
In this scenario, the polls indicate that one candidate already has more than 50% of the vote, so movement of undecided voters wouldn't sway the race [poll movement in the last week could, however].
50% scenario using 538.com
[At 538.com, Nate Silver uses aggregate polling data, weighted for reliability, and adjusted via regression analysis for relevant demographics. Where recent polls are not available, 538.com's model makes assumptions based on polling trends from demographically similar states.]
Obama 286
McCain 142
Battlegrounds (110): AZ, FL, IN, MO, MT, NC, ND, NV, OH, WV
Under the 50% scenario, Obama has already won.
52% Solution
While 50% is a magic number in media analysis, barely over 50% isn't enough to put a state in the bag, since it would only take a very slight swing to tip it the other way. If we make it 52% instead, that adds CO, GA, LA, MN, MS, NH, SD, VA, and WI(!) to the mix, for a projected total of
Obama 235
McCain 114
Battlegrounds (189)
McCain's odds improve slightly in this scenario, which has a fairly high number of battleground votes to contest, but NM and PA are still safe for Obama.
Only 538 provides the internals for this scenario using more than one poll, so let's compare a variety of sites using a different -- more common -- scenario, in which a candidate leads by 10 points in the polls or projections for a particular state.
Double-digit scenario
Why double digits? Not just because it's a nice round number that seems safe, but because there were 7 point swings in 11 states in 2004 between the poll numbers one week out and the election itself. One of those swings cost Kerry Colorado. In Arkansas, which was considered too close to call, Kerry lost by 9.8. Eighteen states (more than a third of them) changed more than 5 points, which is the cutoff many electoral maps use for leaners. There were variances even larger than 10 points in some places (e.g., AL, MA, MS, DC) but none of them changed which way a state was heading, they just changed the margin of a state landslide. So, 10 points it is.
Double-digit lead using 538.com snapshot
Obama 259
McCain 120
Battlegrounds (159): AZ, CO, FL, GA, IN, LA, MO, MT, NC, ND, NM, NV, OH, SD, VA, WV {+ NE 1st & 2nd Congressional Districts}
If you use 538.com projections instead (which expect the race to narrow), PA, MN, and NH get added to the battleground states, as Obama's lead is projected to drop, and LA moves into McCain's safe states. Even though MS moves the McCain column into the prize pool, there's a net gain for him, making it 224 to 123. This is the worst-case scenario for Obama under current poll numbers. The worst case, and he's 46 points from victory, while McCain is more than 3 times as far away, with 147 EV yet to go.
Electoral-Vote.Com [got Kerry/Bush wrong but have since modified their algorithm. Polls are weighted by recency but not by reliability. Polls more than a week old are thrown out if more recent polls are available. No provision for states with split EVs (ME, NE)]
Obama 264
McCain 121
Battlegrounds (153):
[Tossups: MO, MT, ND; Barely Ahead: FL, GA, IN, NC; Weak Lead: AZ, CO, MS, NV, OH, SD, VA, WV]
"Weak" not explained, but probably less than 10%. "Barely" is less than 5%.
Election-Projection.com [Conservative-hosted site accurately predicted Bush in 2004 election, missing by 3 votes with 2 states flipped (IA/WI); uses aggregate of most recent 7 state polls, adjusted by national polls and results of last election; also had good prediction record in 2006 elections]
Obama 255
McCain 112
Battlegrounds [lead of less than 10%] (171): AR(!), AZ, CO, FL, GA, MS, MT, NC, ND, NH, NM, NV, OH, SD, VA, WV
Real Clear Politics.com [couldn't find an explanation of their methodology, but it appears to be a simple average of polls, with partisan-affiliated polls not included; in 2004, RCP did not call sufficient states to name a winner, but had Bush leading in national polls] (with leaners as well as tossups treated as battlegrounds)
Obama 238
McCain 112
Battlegrounds (173): AZ, CO, FL, GA, IN, MO, MT, NC, ND, NM, NV, OH, PA(!), VA, WV
Pollster.com [regression-based trendlines] (leaners as well as tossups treated here as battlegrounds)
Obama 272
McCain 123
Battlegrounds (143): AZ, CO, FL, GA, IN, MO, MS, MT, NC, ND, NM, NV, OH
Pollster.com, with it's refreshingly simple URL, gives Obama the win even if McCain wins all their remaining leaners and toss-ups. No worst-case scenario here using current polling numbers.
Federalreview.com[right-wing site. Has Obama winning 338 to 160.] I'm not bothering to include their numbers, but I mention it for those who like to frequent RedState or FreeRepublic, so you can check out their anti-Obama visuals, including the Obamunist hope w/hammer & sickle logo, and the Obama "thrall" poster.
The Daily Kos Political Scoreboard doesn't show leaners, so we can't question the underlying assumptions but the map is fun to play with, and has time zones to help with watching returns.
http://images.dailykos.com/... for current projection or http://scoreboard.dailykos.com/ to start with 2004 totals
CONCLUSION
If we look at the data from the various sites, we can observe some commonalities.
True Battleground states -- too close to call no matter how you slice it:
FL, IN, MO, MT, NC
Battlegrounds where one candidate is considered safe in some scenarios, but not all, and not comfortably: CO, ND, NM, NV, OH, VA
"Consensus" Battleground Tally:
Obama 259
McCain 157
Battlegrounds 122 (CO, FL, IN, MO, MT, ND, NM, NV, OH, VA)
Possible upsets, considered leaners in some scenarios, but generally projected as at least moderately safe (in bold is upset McCain win): AZ, GA, LA, MN, MS, NE1 (East), NE2 (Omaha), NH, PA, SD, WI (more a Hail Mary than an upset), WV
No longer a battleground, safe for Obama despite 2004
IA, MI, OR
If McCain wins all the close and "semi-safe" battleground states in the consensus tally, he wins 279 to 259 without having to win an upset state, such as PA. The only difference between that result and 2004 is Obama's hold on IA. McCain has 6 ways to achieve a winning combination of states totaling 270 or more electoral votes (vs. Obama's 14 routes to victory), but must win FL, OH, NC, VA, IN, and MO. There are NO must win battleground states for Obama.
Chances of a McCain win per 270towin.com, with consensus battlegrounds in play: less than 1%.
Chances of a McCain win per 538.com: 3.7% (Has fluctuated between 3.3 and 7 over the last week)
Of course, the Cassandras among us will recall that the Rebels blew up the Death Star despite immense odds. But then, the Force was with them. This time it's with us.
Bonus: most likely (but still unlikely) possibility of a 269-269 electoral college tie:
Obama takes Northeast down through MD, upper midwest (IA, IL, MI, MN, WI), West Coast, plus NM and NV.
He would then win in the House.