Friday, September 19, 2008

The Battle Lines are Becoming Clearer

Obama/Biden holds 273-265 Electoral College lead, Dead Heat in Popular Vote Polls

Since my last posting, the Palin bounce has worn off somewhat and the financial crisis has somewhat altered the discussion in the campaign, at least in the short-term. The result has been that Obama/Biden have pulled back at least even in national polls, possibly even up by 1-2% depending on which poll you like. The electoral vote remains unchanged, although I think it is fair to say that the number of battleground states continue to diminish. Overall, even though Mccain did not pick up states, he is in better shape in terms of the electoral map than last week as a few states he was needing to defend as battlegrounds dropped won or off the list. No changes to the map, here are the changes to the battlegrounds:

Key Battlegrounds
None qualify this week

Serious Battlegrounds
Colorado -- unchanged from last week -- I continue to contend that this state will be the key to the race (more discussion on this below)

Substantial Battlegrounds

New Mexico -- unchanged form last week -- Obama holds a small but consistent lead

Somewhat Battlegrounds

Ohio -- demoted from serious battleground -- this continues to solidify for Mccain -- losing her would essentially rule out a national win for him
Pennsylvania -- promoted from potential to somewhat -- a couple of polls have showed it tied -- this is like Ohio for Mccain, Obama can't afford to lose here


Nevada -- demoted all the way from key to fringe -- Mccain has been strengthening here for the last couple of weeks -- relatively small African-American population makes this a hard win for Obama -- he will need high hispanic support and turnout
Virginia -- demoted from somewhat to fringe -- it is still close, but Obama's dream of picking this up seems less and less likely -- he will need massive wins in Arlington, Alexandria, Fairfax and Loudon Counties as well as the City of Richmond to pull it off
North Dakota -- demoted from somewhat to fringe -- I can't seriously see this one not going red
Minnesota -- promoted from potential battleground to fringe battleground -- polls are remarkably tight -- location of the Republican convention may have provided a bounce to Mccain -- a loss here would be devastating for Obama

Note: Florida, Montana, Indiana and New Hampshire have all been dropped as battlegrounds. I have elected to include all 4 as potential battlegrounds as polling is still relatively close.

Potential Battlegrounds
Michigan -- I kept this on the list as polls are relative close, although I can't see Obama losing here unless Mccain wins huge. Economic issues should help Obama here as Detroit obviously continues to be hit hard.
Missouri -- basically the same story as Michigan in reverse -- if Obama wins here, it'll be a blue sweep
North Carolina -- I thought we would be done talking about NC by this time, but the polls are remarkably inconsistent, so I'm not sure what the state of the race is here. Mccain clearly leads, but could be by 1% or 20% depending on who you believe.
New Hampshire -- I may drop this one soon as Obama appears to have solidifed a strong lead here but this state has always loved Mccain and is still far more conservative than the rest of New England.
Indiana -- how can this still be within 5%? The only reasonable explanation is tough economic times and the effect of the Chicago suburbs. Tough to believe Obama could win this super-red state he couldn't even win in the primaries.
Montana -- like North Carolina, a state with very inconsistent poll data. Only 3 electoral votes, so they must have called everyone in the state by now with all the polls that have been taken.
Florida -- fading as a battleground as Obama appears to be running behind where both Gore and Kerry were with older Jewish voters. Would be a longshot but also a walk-off homer if Obama could score an upset here.

The North Carolina / Montana Effect and the Truth about Polling
Within the past couple of weeks, there have been broad variances in the poll results from different organizations. Nowhere is that more clear than in North Carolina and Montana. Polls in North Carolina have shown the gap between Mccain and Obama anywhere between 1 and 20 points and Montana anywhere from 3 to 15 points.

What gives?

I don't know for sure, but here are a few thoughts:
(1) Margins of Error
Possibly the most misunderstood aspect of polling data is a poll's "margin of error". This term is in itself a misnomer. What is normally referred to as "margin of error" is, in fact, a calculated number known to statisticians as a Confidence Interval. The principle works like this -- the more samples that I take out of a population, the more confident I can be that the sample I'm taking is relatively representative of the population at large. Think about 100 beans in a jar, some red, some blue. If I pull 1 bean and it happens to be blue, it doesn't tell me much about the contents of the jar - it could be 55% red or 55% blue and it would be a reasonable outcome for me to pull 1 blue bean on any given pull. Now, let's say I pull 10 beans. If 8 of 10 are red, I'm pretty confident that there are more red beans than blue beans. Of course, I'm not 100% confident, I could have just happened to pull a disproportionate number of red beans. The same basic principles works with polls. When a polling firm calls 1,000 people in North Carolina, they THINK they have a representative sample, but it is POSSIBLE that they just happened to call people who are disproportionately pro-Obama or pro-Mccain. Most polls that publish a 3% Margin of Error are expressing a 90% confidence interval. That is, based on the sample size, they are 90% confident that the individual results are within 3% of the actual results. 10% of the time, the error will be higher. By the way, because the 3% applies to the individual results, the effective range of the poll is actually 6%. In other words, if Mccain leads Obama by 6% in North Carolina, within the 90% confidence interval, it could be tied or Mccain could lead by 12% if the margin of error of the poll is 3%. Oh, and by the way, the 10% of the time that the actual results fall outside of the confidence interval, Mccain could lead by more than 12% or actually be behind. Confused yet?

(2) Sample Error

The statistics behind polling assume that the pollster is getting a truly random sample. That is, of the people who are going to vote on election day, he is equally likely to talk to any of them. Of course, pollsters can't do this because a. they don't know for sure who will vote on election day, b. not everyone will speak to pollsters and c. 10% of voters no longer have home phones. Pollster recognize these problems and attempt to "normalize" their data to adjust for the people they missed. In other words, if they believe that women will represent 52% of the votes cast in a state but their poll results show that only 35% of the people that they talked to are women, they will adjust their results to give a disproportionate weight to the women they did talk to.

The problem is that nobody knows for sure who will vote and nobody can tell for sure that the people that they did talk to are more broadly representative of the groups (e.g. under 25 voters without home phones may have different voting patterns than under 25 voters with cell phones.)

All of this makes accurate polling a very complex operation and can lead to wide variability in poll results depending on the methodology applied

(3) Inclusion / Exclusion of Third Party Candidates
A CNN/Time poll I saw recently showed a 2-3% swing in favor of Obama when third party candidates were included on the list of options. This didn't inherently make any sense to me, but apparently both Nader and Barr are stealing more from Mccain than from Obama (or maybe just mentioning Nader makes Democrats unite around Obama.) Regardless, it is hard to tell if this will play out in the voting booth, but it accounts for some of the differnece in some polls.

(4) Registered vs. Likely Voters
Some polls attempt only to survey the total registered voting population. Other polls attempt through a series of questions and historical patterns to determine which voters will actually show up at the polls on election day. The problem with this is that it is difficult in a given election (especially one as dynamic as this one) to know who will actually come to vote.

All told, polls are a valuable tool still, but keep in mind they are just one data point. That is why it is best to consider all the polling data as well as historical and national trends to figure out who is going to win a state. Amazingly, with the glaring exception of New Hampshire, polling was by and large very accurate during the primary season.

Colorado is the Key
I can forsee 3 possible scenarios in this election:
#1 Obama breaks out and takes a substantial lead -- I still believe that it is possible given the current political climate and unpopularity of the president that Obama could be leading by 5-7% a month from now. This would lead to a crushing electoral vote, with Obama carrying at least 350 electoral votes.
#2 Mccain finds a way to win in Pennsylvania -- if Obama can peel off enough Hillary voters and win big in places like Allentown and Scranton, he could effectively end the race early in the night. It is hard to draw up a realistic map in which Obama loses PA but wins the election
#3 It all comes down to Colorado -- in map after map, Colorado holds the key to the election. Makes the Democrats look pretty smart for holding the convention in Denver. Put on an extra cup of coffee on election night while they count the ballots down there.

269 to Win
We talk about a candidate needing 270 electoral votes to win, but in reality, John Mccain needs 270 and Barak Obama needs only 269. The reason? A 269-269 tie sends the election to the new House of Representatives with each state having one vote. Democrats will likely control 35-36 state delegations in the new congress. Will all the Democrats in the House necessarily vote for Obama if their state voted against him? YOU BET THEY WILL! Survival instinct dictates that they will support their parties candidate, particularly after all the controversy in 2000 and 2004. Besides, the whole point of it going to the House in a tie is so that they can use their own judgment, not simply repeat the same votes that got us to the tie in the first place. Republicans would do the same except they can't.

By the way, the new Senate picks the VP by straight vote, so Biden is in good shape as well.

269-269 is a real possibility. Look at my current map. Say Mccain picks up Colorado and Obama picks up Nevada. Voila! 269-269. Or say Mccain picks up New Hampshire -- poof! Back at 269-269. And you thought Bush v. Gore was controversial.

The Dynamics are Complex
Just a few thoughts to ponder as we get ready for the debates:
#1 How will all these new voters the Obama camp is registering impact the election? Could they make the current polls totally wrong?
#2 Are some white voters saying they are going to vote for Obama but will have a last second (or pre-meditated) change of heart in the voting booth? Don't rule it out.
#3 Is there an October surprise lurking out there for either candidate?

One week until the first debate. Could shake things up or solidify the battleground -- we shall see.

No comments: