Imagine someone hands you a sealed box which, they say, contains a lot of money. You take the box home, open it, and find three dollars inside. You can now ask yourself one of two questions: "Where did all that money go?" or "I wonder how much money was really in that box in the first place?"
At the beginning of this year, we were given a stack of boxes marked "Democratic Primaries -- Property of Howard Dean" that, we were told, had a lot of votes in them. Over the last few weeks, we've opened several of those boxes, and none of them have contained the votes we were told to expect. And over and over, people are asking "Where did all those votes go?" without asking whether the votes were ever there in the first place.
The press is bad about this. The NY Times blithely suggested that the Dean campaign started "running out of steam" in January, as if facing actual voters was just one factor among many, rather than the main event. Dean supporters are bad about this as well -- just recently, in a comment left on
my last Dean post, someone said "Those poll numbers were samples of the votes of the people, and they were accurate right up to the Iowa caucus (i.e., they started dropping before caucus night). Dean’s popularity wasn’t imagined. He would have won if the caucus was held a week or two earlier."
This badly misconstrues what happened at the polls. Dean's numbers didn't start going down on a particular calendar date, they started going down a couple of weeks before the voters had to make up their mind.
We know from New Hampshire that Dean's support evaporated in last few weeks before the campaign, and yet he beat Kerry among voters who had made up their minds more than a month in advance. This means that, by definition, a sizable number of New Hampshire voters told pollsters they would vote for Dean when they hadn't actually made up their minds. Had the vote happened sooner, the voter's actually deciding would have happened sooner as well.
Contrary to the earlier comment, poll numbers are not a sample of votes. Polls have become such accurate proxies for elections that we've forgotten that answering a pollster and casting a vote are very different activities -- polls are iterative and non-binding, while a vote is a single binding event. A poll differs from an election, in other words, in that it is public signaling without affecting any actual outcome.
The question pollsters want respondents to hear is "If you were to regard this poll question as your vote, who do you favor?" It may be that decades ago, in an age where polls were rare and respect for authority was reflexive, respondents tried their best to answer that question. Today, however, is different. Today respondents hear "Knowing that this question is not binding, and that it will be a signal to people who want to make decisions based on your answer, and understanding that if you say 'Not sure' you are largely nullifying your ability to send such a signal, who are you going to tell me you favor?"
And what seems to have happened with Dean is that the people answering the polls signaled interest in Dean, without knowing if they'd vote for him, because they liked what he seemed to stand for -- a tough-minded Democrat willing to challenge the President on a host of important issues. And the effect of that public signaling seems to have been two-fold -- first, it convinced a lot of us that there were a lot of votes in those boxes (some people still believe that those votes used to be there and somehow leaked away, though the enormity and totality of Dean's losses makes that a hard view to support.)
The other effect is that it seems to have arranged a partial transplant of both issues and spine from Dean to Kerry. I pointed to a Chris Sullentrop article last July on
the threat the Web poses to the Dean campaign, which suggested that the Dean campaign was in danger of letting its bottom-up organizational style attribute to the candidate views he didn't actually hold. (He was, after all, a fiscally moderate Governor with a pro-gun record.)
This "Dean as projection of the voter's will" notion, however, may have a silver lining for the Democrats. Because polling is not the same as voting, and because the voters didn't know much about Dean other than that he stood for "the Democratic wing of the Democratic party", he may have been the ideal way of surfacing the issues and strategies the voter's would respond to, while committing them to no particular candidate. If Bush loses in November, Dean will have contributed in no small part, not by winning votes, but by being a useful screen for the projection of public signals.
1. Shannon Clark on February 17, 2004 2:45 PM writes...
This raises the following thought:
What if polls and a campaign were to start with collecting the issues and positions of a large block of voters.
And then (and only then) match them to a candidate that "fit" them well?
i.e. could social networks, online & offline organizing, polls, etc be used to capture the definition of a successful candidate - who would then be found amongst a pool of potential candidates (for example elected officials of the "right" party or currently independant with track records that match up to the issues and positions identified"
On some level this is a bit of what goes on behind the scenes today, the people with the purse strings of a given party do some "vetting" of candidates and suggest to some that they consider running, to others that while they are free to run the look unlikely to win, etc.
But as the Dean campaign and MoveOn have shown, large numbers of people properly motivated can represent a very potent fundraising and activity conducting base - could this be used even before a candidate was found? (potentially arising from the body politic itself?)
Just a thought.
Shannon
Permalink to Comment2. Lucas on February 17, 2004 6:26 PM writes...
Shannon, you have hit the nail on the head. People feel disenfranchised not because they believe their vote doesn't count but because of the odious choices offered to them. Really this is the best way Internet activism can have an effect on the political process. It would at once emphasize issues over the horserace and better select a candidate who people believed genuinely had their interests at heart.
Programmatically it is an application of clustering (identifying constituancies based on groupings of similar stances on various issues) and user-centric collaborative filtering (to match candidates to these clusters).
The second part that is necessary to affect change is a contract between the candidate and those citizens who elevated them into the spotlight, similar to Newt Gingrich's much derided Contract With America. The contract would be based on the results of the stances contained in one of the clusters. (Without this citizens must rely soley on their less reliable instincts of personality evaluation.)
Permalink to Comment3. JayT on February 17, 2004 10:49 PM writes...
I'm sorry, but these are some sick comments.
Yeah, that's what we just DID have, with Dean. Cookie-cutter candidates, made-to-order by the electorate. There's a VERY GOOD reason we have a representative democracy, how's it go?... "The vagaries and wild swings of the populace".. something like??
Mr./Dr? Shirky, "the enormity and totality of Dean’s losses makes that a hard view to support" because you wanted him to win, in the first place.
"If Bush loses in November, Dean will have contributed in no small part, not by winning votes, but by being a useful screen for the projection of public signals."
Well, back in 2000 I heard from my then-Wife who heard one-a the "talking heads" on TV mention history of candidates who won the popular vote but were voted out by the Electoral College. So that's, what I call, a semi-reputable piece of information. So I checked out the facts, on the first time it happened in 1888, I think, and sure 'nuff: The next election the candidate was elected by a large margin. (So now I assume same of other 2.)
The "revenge" of the DNC and the American People, on Al Gore, was in not raising a big enough issue of the Electoral College.. never even came to a vote... (Then 9-11, and a Republican shift in the Congress, because who wants to admit a mistake or two HAD been made.)
So, barring Dean, I would say it unlikely that President Bush would have gotten elected. Because of a lot of reasons, and history does tend to repeat... Most unlikely, but now I'd put the odds about even, depending mostly on 6/30 situ in Iraq, would be my guess.
Dean and Joe Trippi may have raised some issues, and awareness and activism, no doubt... But if a Democrat gets elected, it will be in SPITE of them, a LOT more than because of them.
I'm sorry, Dr. Shirky, but I don't see the point of making up a silver lining where one doesn't exist, because it will skew the interpretation of the data, now and/or in the future.
The voter's made a better choice than the Deaniacs, because why would anyone WANT to be subliminally? known as a maniac who is for Dean???
Permalink to Comment4. Arnold Kling on February 18, 2004 12:59 PM writes...
I'm repeating myself, but my view is this:
Dean's support as of early January was a mile wide and an inch deep. Kind of like Kerry's right now.
Dean and his supporters came into Iowa with an attitude that turned off Iowans. His support melted.
The most ardent Dean supporters, who talk most fervently about this being a campaign "of the people," are totally out of touch with the people. They live in an echo chamber.
The people don't want to be saved by Dean and his holy rollers. They want a candidate who doesn't scare them or offend them.
I'm not saying that the people know best. I'm just saying that's reality.
Permalink to Comment5. Seth Finkelstein on February 18, 2004 2:50 PM writes...
Agreed.
Why is this hard or complex? Polls aren't votes. Buzz isn't action.
It doesn't have to be social signaling in game-theoretic emergent herd-influencing behavior.
If a pollster asks, "Who are you likely to vote for?", weeks before the election - people's answer's aren't final. They don't even necessarily represent any thought at all. It's something said to a caller over a quick phone conversation, possibly even the first name to pop into one's head (which is why press coverage can give a distorted result). There's absolutely no reason to *project* early results into final votes, especially if much intervenes in that time, which is the key reasoning step here.
Permalink to Comment6. tunesmith on February 18, 2004 8:40 PM writes...
This concept isn't new in the political press. There are plenty of stories of voters who supported Dean but "when it came right down to it" voted differently. The point isn't that Dean's support was illusory. It was that the voters judged that this period of time called for using different standards. Basically, Dean represented something they wanted, but that didn't fit with what they diagnosed was most needed at this time.
What would prove me wrong? If support for Dean's movement fizzles after Kerry beats Bush. I highly doubt that will happen, though - there will be a large appetite for its continuance, from more people than just the people that actually did vote for Dean.
Maybe at another time the voting population will be ready for what Dean represents. But 2004 is about practicality more than ideals.
Permalink to Comment7. tib on February 22, 2004 4:51 PM writes...
Yes, polls are signals, hence the rapidity with which the other candidates adopted Dean's style.
Primary voting preference has always been volatile. Gallup began asking about nomination preference in 1936, and voter preference swings wildly prior to election day in every competitive primary since then. As Samuel Popkin points out in "The Reasoning Voter" this is because voters decide late in primaries, and the larger number of candidates typical in a primary make for a more complex decision with more opportunities for strategic voting (Kucinich voters go for Edwards in IA).
Permalink to Comment