October 22, 2008

Don't Give Up the Ship Until the Last Fat Lady Is Hung

Hatched by Dafydd

Take a look at these polls listed at Real Clear Politics, as of 11:40 am PST, October 22nd, 2008; each poll was released today and covers either through October 21st or through October 20th. All trends are based on the previous poll (yesterday's unless marked) and are relative to Barack H. Obama... so +2 means 2 points better for Obama, -2 means 2 points better for John S. McCain:

  • Zogby: Obama +10 (trend +2)
  • AP: Obama +1 (trend -6 from October 1st)
  • NBC/WSJ: Obama +10 (trend +4 from October 6th)
  • Battleground: Obama +2 (trend +1)
  • Fox News: Obama +9 (trend +2 from October 10th)
  • IBD: Obama +4 (trend -2)
  • Ipsos: Obama +8 (trend unknown -- last poll was registereds, not likelies -- probably positive)
  • Gallup trad: Obama +5 (trend -2)
  • Rasmussen: Obama +6 (trend +2)
  • Hotline: Obama +5 (trend -2 from October 10th)

(Polls in blue trended towards Obama, those in black trended towards McCain.)

Not only are the polls all over the place -- Zogby has Obama up 10 points, AP has Obama up only one point? -- but even the trends are all over the place, from 4 points towards Obama to 6 points towards McCain.

This is a near perfect illustration of how different respondent pools, order of questions asked, and turnout assumptions all affect poll results. We cannot single out any particular poll in advance and declare that poll to be the "correct" one, while the others are more or less wrong. We simply don't know today which poll will prove to be prophetic of election day.

Obviously, Zogby and AP cannot both be accurate, but which should we believe? What Real Clear Politics does is simply take the mean average of the polls: They add up all the recent poll numbers for Obama and divide by the number of polls, do the same for McCain, and compare them. But the standard deviation here is hellish; it's like the old joke...

Three statisticians go hunting. They see a deer, and two of them fire simultaneously. The first misses 12 feet to the right, the second misses 12 feet to the left -- and the third whoops, "On average, we hit the buck dead center!"

It's useless to average a +10 poll and a +1 poll to say that on average, Obama is 5.5% ahead. It's reasonable to suppose that Obama is "really" 8-10 points up; but it's just as reasonable, and just as accurate, to suppose that he is "really" only 2-3 points up. And if the latter turns out to be the "correct" figure -- that is, if the turnout assumptions that produced the lower figures more accurately match what happens on November 4th than the assumptions that produced the higher lead -- then John McCain has a very strong chance to win the election.

Democratic turnout will certainly be a bigger percent of the electorate than in 2000 and 2004, but how much bigger? Let's label the two scenarios illustrated by the polling above "big" Democratic turnout and "tsunami"-sized Democratic turnout. A projected big Democratic turnout yields a 2-3 point current advantage for Obama, while a projected tsunami turnout for the Democrats yields an 8-10 point advantage. Since we don't know at this point whether Democratic turnout will be merely big or a tsunami-like tidal wave, we cannot begin to guess how far ahead Obama is... or even whether he is catchable.

If it turns out to be tsunami, then nothing McCain does can change the outcome: He will lose, end of analysis. So let's look at the other scenario exclusively. Here is a very useful tool, the Real Clear Politics' "Create your own map" facility. You can click on different states, set them to either Obama or McCain, and see how that affects the final electoral count.

Currently, counting leaners, the RCP map shows Obama/Biden with 286 electoral votes, McCain/Palin with 160, and 92 votes are contained in states that are toss ups, meaning leaning one way or the other by less than five points (in fact, all but one lean by less than three points).

We're assuming, for sake of analysis, that the Democratic turnout is big, but not tsunami. If that is the case, then the structural Democratic advantage inherent in most polling means that McCain is likely to win all the states currently listed as toss ups: Florida, Indiana, Missouri, Nevada, North Carolina, North Dakota, Ohio -- all of which were won by George W. Bush in 2004 and 2000. So let's go ahead and change them on the map from "toss up" to "leaning McCain."

When we do that, the electoral vote becomes Obama/Biden 286, McCain/Palin 252. So what states does McCain need to switch to win the election?

The two that spring readily to mind are both Bush states from 2004 and 2000: Colorado and Virginia. I frankly disbelieve the polling showing Obama ahead by 5.4% in Colorado and by 6.8% in Virginia. But even if we accept those numbers, neither is very significant: Colorado is just barely out of the toss up category, and even Virginia is subject to assumptional poll fluctuation (Rassmussen has Obama up by 10, but Mason-Dixon has Obama up by only 2). Again, if we're talking the big scenario, not the tsunami scenario, then these two states are very winnable.

If McCain wins them -- taking not one single blue state, and giving up the former (slightly) red state of Iowa, which has become deep blue in the last four years -- then John McCain wins the election by 274 to 264. In fact, he could even lose either Montana or North Dakota and still win (barely).

Curious sidebar: If McCain wins Colorado and Virginia, plus all the toss ups except Nevada, then we have a 269-269 tie; the race would be decided by the House of Representatives, with each delegation getting one vote -- and that means Obama wins, because Democrats currently control 27 state delegations in the House, while Republicans control only 21; 2 are split... but even if they both break for McCain, he still loses by 27 to 23.

Joe Biden is almost certain to be chosen as vice president in this case, because the Twelfth Amendment appears to leave the VP selection to the Senate on an ordinary majority vote; the Senate currently comprises 49 Democrats, 49 Republicans, and two Independents who caucus with the Democrats; but even if Joe Lieberman votes for Sarah Palin, there will certainly be more Democrats in the Senate in the new 111th Congress, which would do the voting in such a case.

This is why it's ridiculous to panic, despair, and resign ourselves to President Obama: Everything, even the winner, still depends upon which turnout assumption we pick; Each outcome still has support in the polls. Neither outcome is the overwhelming favorite.

So as I've said many times, it's time to put on our manly gowns, gird our loins, and pull up our socks. Let's go out there and win one for the old nipper!

Hatched by Dafydd on this day, October 22, 2008, at the time of 5:10 PM

Trackback Pings

TrackBack URL for this hissing: http://biglizards.net/mt3.36/earendiltrack.cgi/3297

Listed below are links to weblogs that reference Don't Give Up the Ship Until the Last Fat Lady Is Hung:

» Some Interesting Polling Figures and Map Games from Big Lizards
Currently -- before most of the October 26th polling trickles in -- Barack H. Obama is ahead in the Real Clear Politics average by 7.6%; but that includes some whoppers (in both senses of the word) from several days ago,... [Read More]

Tracked on October 27, 2008 7:13 AM

» ...And the Elite Media Are NOT the "Fat Ladies" We Mean from Big Lizards
I titled an earlier post Don't Give Up the Ship Until the Last Fat Lady Is Hung. I want to make it clear that each and every one of the drive-by media anchors is going to try to become a... [Read More]

Tracked on October 30, 2008 5:28 PM


The following hissed in response by: antimedia

It's also foolish to assume that a high Democrat turnout is good for Obama. There are a lot of angry Democrats who are voting (and some have already voted) for McCain/Palin. There are three reasons for this; Hillary supporters who believe she was cheated out of the nomination, Joe the Plumbers who were insulted by Barack's treatment of Joe and Democrats who are opposed to socialism and believe Obama is a true Marxist (as do I.)

The above hissed in response by: antimedia [TypeKey Profile Page] at October 22, 2008 7:37 PM

The following hissed in response by: Dick E


I’m certainly ready to gird my socks and pull up my loins.

But I have a basic question about public opinion polling: What public benefit does it provide?

I know you’re an expert on the subject, so I’m treading on pretty thin gruel here, but it seems that polling is like throwing darts, dressed up in scientific nomenclature. Precision, confidence, sample size, etc. are perfectly valid statistical concepts. But how can results as wildly disparate as those above be statistically valid? And even if you argue for statistical validity, how are they useful to the public?

Unfortunately, opinion polling involves serious problems, some of which you have discussed previously. For example, different questions used by pollsters and the tendency of people to interpret questions differently or to just plain lie. Then there’s the mix of respondents, e.g. Democrats vs. Republicans. Attempts to adjust poll results for these factors involve judgment and thus reduce, rather than enhance, statistical validity.

But even if the methodology were unassailable (which, obviously, I do not acknowledge, but would defer to serpentine expertise), what pubic purpose is served by these polls? What should voters do with the information? Change their votes to be more in or out of tune with the majority? I certainly hope not. Use it in political discussions? OK, but again, what purpose is served other than trying to influence people to vote like or unlike others?

Don’t get me wrong: I don’t propose to ban/limit/regulate polling. Freedom of speech and the press are too precious to squander. Plus there is, of course, money to be made in them thar polls. And certain pundits -- including some green and scaly ones -- need grist for the commentary mill. I just think we pay far too much attention and lend far too much credence to these polls.

(Of course, political parties and candidates can use polling information, however imperfect, to allocate campaign resources. That’s perfectly legitimate.)

The above hissed in response by: Dick E [TypeKey Profile Page] at October 22, 2008 11:08 PM

The following hissed in response by: Dafydd ab Hugh

Dick E:

But even if the methodology were unassailable (which, obviously, I do not acknowledge, but would defer to serpentine expertise), what pubic [sic] purpose is served by these polls?



The above hissed in response by: Dafydd ab Hugh [TypeKey Profile Page] at October 23, 2008 5:20 AM

The following hissed in response by: Pam

Yes, things do not look the best right now, but we still have 12 days to Turn out, Turn out, Turn out the vote.

Way to fire up the troops,Dafydd!

Folks if we turn out like we did in 04, and why shouldn't we, we've got Sarah Palin on our side this time, WE WILL WIN!


The above hissed in response by: Pam [TypeKey Profile Page] at October 23, 2008 7:10 AM

The following hissed in response by: Dick E


Sorry, I was still busy girding my loins, and it just slipped out.

The above hissed in response by: Dick E [TypeKey Profile Page] at October 23, 2008 11:37 AM

The following hissed in response by: Xpressions

Again, these polls are not votes. If we all believe our vote won't matter, that's the goal of the socialist illuminati. The fact remains, you must go cast your vote, because your vote matters. Just to comment on your vote counting, it does, but it's your money that will be spread, so you count!

The above hissed in response by: Xpressions [TypeKey Profile Page] at October 24, 2008 7:24 AM

Post a comment

Thanks for hissing in, . Now you can slither in with a comment, o wise. (sign out)

(If you haven't hissed a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Hang loose; don't shed your skin!)

Remember me unto the end of days?

© 2005-2009 by Dafydd ab Hugh - All Rights Reserved