On Sat, 24 Jul 2004, Michael Perelman wrote:
> I also have trouble figuring out the widely different polling
> results.... Are the swing voters so volitile? Are the pollsters asking
> different questions? Or should we just ignore the horse race?
They are basically impossible to figure out from the papers. 98% of the summaries you read are garbage. Polls have definite limitations, but there is a right way and a wrong way to get at what they do have to offer. The right way is not to pay attention to individual results, but to medium term trends, just like with economic indicators. The second important thing is that these trends can only be charted for the same poll -- you can't compare between polls without heavy qualification.
The papers basically do exactly the opposite: they pay almost exclusive attention to the immediate number and they compare across polls all the time. Then on top of that is a good deal of downright deviousness that never makes it into the papers as an issue.
Lastly -- and perhaps most importantly -- you can't predict the trend in the horse race by tracking the horse race. *Especially* this far in advance, when you are forced to use registered voter models rather than likely voter models. (A month or two out is little different.)
The electoral horse race is a little like the price of a currency between two countries. Sometimes it stays perfectly steady while the underlying factors that should affect it (like the comparative inflation rates, interest rates, growth rates, equity returns, trade and fiscal position, etc.) are clearly going negative. What people usually conclude from that is not that the underlying trends don't matter anymore, but that the currency will fall -- the only question is when.
That is essentially the position Bush is when it comes to the polls. The underlying trends in this case are not causal trends (as they are with the currency) but more in the nature of weekly leading indicator trends that have an excellent predictive record: net balance of favorability, right track/wrong track question, and approval/disapproval of various areas of policy, along with ratings of which area of policy will most affect their vote.
Bush's readings in on these underlying trends -- every damn one of them -- have been trending drastically south for months. And they are much worse among Independent voters, and worst of all among independent voters in battleground states.
> Most people's preferences are reportedly set in stone.
That's not true. It feels that way, but it's not.
To start with, just as a poll result, according to the latest CBS/NYT poll:
20% of all voters say it's too early for them to have finally made up their mind.
In the larger picture, in terms of party ID, the country is roughly divided into 30% Republican, 40% Democrat, and 30% Independent. The exact numbers from the poll cited above are 29% Repug, 37% Dem, 30% Independent, and 4% other.
(BTW, just in passing, if you include leaners -- if you exclude Independent as a category -- the same poll breaks down as 40 Repug, 54% Dem and 6% Other.)
Now, as with financial indicators, past results are no guarantee of future results. Plus the time is limited. Currency markets often stay in what looks like an out of whack place for years. In a political market, if they stay that way six months, it can be all over.
And lastly, political markets differ in that causative events that chance everything can be created through a successful political attack. In that sense, reading polls are more like tracking a war than an economic cycle -- it's not won until it's won.
But with all those very big qualifications, on current trends, Kerry wins if the election is held tomorrow, and if the trends were to continue, he would win by even more in the future. And that seems to be visible in their actions at this point. Bush seems to be shoring up his base and Kerry seems to be ambitiously reaching beyond it.
If you really care about these poll issues (and there is no reason why you should), I highly recommend subscribing to Ruy Teixeira's email newsletter, Public Opinion Watch, at the Century Foundation (http://www.tcf.org) Each week he covers all the major polls of the preceding week and analyzes them the way they ought to be analyzed as well as proffering explanations for discrepencies and pointing out where the news summaries are wronger or more devious than they usually are.
Michael