try another color:
try another fontsize: 60% 70% 80% 90%
Overdetermined

The Undecided Fallacy

Sat, 10/25/2008 - 16:06 — Student Redux

I've heard a lot of sturm und drang over the persistence of undecided voters in the closing weeks of this campaign.  "Why can't Obama close the deal," is a frequent media meme, even on MSNBC.  Every time I hear this question, I'm more than a little surprised.  Why?

Because Obama closed the deal weeks ago.

There's more...

Now as an Obama supporter, believe me, I'm not advocating complacency.  There's more than a week left in this race, and we can't afford to let up for a minute.  We've seen what happens when you give Republicans an inch.

But let's take a look at Obama's support and McCain's support over the last month, roughly since the onset of the financial crisis.  On Sept. 15th, the RealClearPolitics poll average for the race stood as follows:

Obama: 44.7%
McCain: 46.3%
Undecided: 9%

Where is RCP today?

Obama: 50.4%
McCain: 42.4%
Undecided: 7.2%

Okay, now here's the key point.  Let's look at NET CHANGE:

Obama: +5.7%
McCain: -3.9%
Undecided: -1.8%

On Sept. 15th, 9% of the electorate was undecided.  Of those who were undecided AT THAT TIME, 60% now support Obama.  What's clouding our vision is the fact that, at the same time, McCain has LOST a significant fraction of his own support, increasing the size of the undecided voter bloc.

When the poll numbers shift, we aren't seeing McCain supporters suddenly break and go for Obama.  We're seeing Obama win undecideds, at the same time that McCain supporters reconsider their decision to vote for him.  So, in fact, the pool of undecided voters has almost certainly become increasingly hostile toward Obama - because Obama has already solidified the support of anyone leaning his way.  What's left is a pool of disaffected McCain supporters - a pool Obama continues to win, amazingly.

We saw this in the dial polling during the debate.  As much as I dislike dial polling, it was pretty evident that the dial polls started out very pro-Obama in the first debate, and by the third, the dial polls had lost most of their pro-Obama skew.  Why?  Because CNN recruits undecided voters for their dial polls, and because Obama was sealing the deal with anyone who leaned his way.  It became harder and harder to find pro-Obama undecideds in the pool of undecided voters, and easier and easier to find undecideds with a little bit of a McCain slant (some of whom, at the time of the first debate, probably even counted themselves as McCain supporters).

This media meme that Obama can't close the deal has stood unchallenged for too long.  Obama has been closing the deal for weeks.  The candidate who can't close the deal - the candidate who's actually watching "the deal" slip away from him day-by-day - is John McCain.  Obama is reliably over 50% in almost every major poll.  McCain, despite getting a huge bump for the St. Paul convention and the Sarah Palin pick, is at roughly the same level of support he had during the Democratic National Convention.

One candidate has been winning more and more voters for weeks.  One candidate hasn't seen an ounce of net improvement since the start of the general election.  Who on earth can seriously ask why Barack Obama hasn't closed the deal?

Comments

While this is certainly one possible modus

Mon, 10/27/2008 - 14:48 — ukben

There are other potential explanations. I think it's important outline any of the possible reasons people may respond with undecided. Just off the top of my head, we could further classify undecideds into:

1) Truly undecided voters. Only when they are in the booth will they decide. These voters, for whatever reason, refuse to make up their mind until election day.

2) McCain leaners. Comprised of both people that were voting for McCain and are now disaffected, or traditional conservatives that aren't quite sold on McCain.

3) Obama leaners: Comprised of both people that were voting for McCain and are now considering Obama.

4) Closet supporters who have made up their mind, but won't tell pollsters.

One addiitonal counter to this argument is that the focus groups tended to come up in favor of Obama. Now these I find unscientific, but the same can be said for the dial groups. The sample size is small, and consequently squirrelly IMHO.

I am beginning to wonder whether we should ask about always support. What I mean here is have you always supported who you currently support etc. If we are obsessed with trends then the cross-over question seems a biggie to ask. That would be an importantly informative criteria in the context of mapping trends. I think there's a lot of room for moving the sets of questions.

Finally, my read on the media is that this is not going to be a close contest, really. But we've still got a week to go.

Well...

Mon, 10/27/2008 - 21:25 — Dirty D

UKBen,

While it may be the case that :

UNDECIDED=TRULY+LEAN_OBAMA+LEAN_MCCAIN+CLOSET,

the simple case is that as we get closer to the election, we should see the first three categories diminish as they convert to one choice or the other.  At the very least, TRULY should diminish in magnitude, and I think that you'll movement from the leaners into full fledged support, whether openly or as a member of CLOSET.

The point remains that as you move closer and closer to the election, not only does the magnitude of UNDECIDED shrink, but it shrinks due to loss of members, which means that it's fundamental nature changes.  When you are polling undecided voters in October, you are polling a group that's very, very different than  undecideds in August.

The problem with asking about always supporting a candidate is that, well, people lie.  Few people like being seen as Johnny Come Lately's, and they'll either intentionally or unintentionally misreport what happened.  It's a really fascinating phenomenon - people retcon their memories to suit their current beliefs.   There's no really good way to measure to what degree this happens, either. Maybe, when polling, we could preserve all the undecides from one wave and fold them into the sample for the next.  That way, we can measure this.

Dirty D

 

The funny thing is that I thought that this was self-evident.

Sun, 10/26/2008 - 21:48 — Dirty D

I don't say this to denigrate your point, and it was very well laid out, but it did seem self-evident to me.  It seems to me that likelihood of susceptibility to a candidate's message is directly proportional to the distance from the election.  The farther out the election is, the more likely you are to be swayed by the candidates' messages.

I do think that it's poor journalism to not note this, and to pretend that an undecided voter from January is identical to an undecided voter from October.  Actually, we have a category on this site to describe posts like this - you may want to go back, edit and add the categoriy "Bad Journalism".

Dirty D

Good call

Sun, 10/26/2008 - 22:03 — Student Redux

Edited.  And I agree, it should be self-evident.

Sadly, I don't think that's the case for a lot of people, or we wouldn't be seeing the meme continue to pop up even now.  I suppose it's one of those things - as soon as you think about the question, the answer is staring you in the face.  But a lot of people, I think, fail to even think about the question.  They just accept that undecideds are undecideds are undecideds.