Two weeks from Election Day, it’s now time for political junkies to begin thinking about the dynamics of the Big Day. And right on cue, David Paul Kuhn of Politico has a very instructive primer on changes in the network-sponsored exit poll operation for November 4.
As you probably recall, the early exits in 2004 were a misleading mess. Leaked in every direction, they showed John Kerry ahead in several key battleground states, most notably Ohio. It later transpired that sampling errors–especially an oversampling of two pro-Democratic voter groups, young people and women–skewed the exits crucially (I can personally remember noticing that exit polls from SC showed 60% of the electorate as female, which I should have known made the whole thing suspect).
With younger voters in particular leaning so heavily towards Barack Obama, Kuhn reports that the nets and the Edison-Mitofsky organization that actually runs exit polling are extremely focused on sampling bias. The fact that during this year’s primaries early exits almost always overestimated Obama’s ultimate vote is a data point as well. Aside from the age and gender composition of exit poll respondents, it’s been observed over the years that very enthusiastic voters tend to get oversampled, for the simple reason that they are eager to participate.
So what are the exit wizards doing to improve their accuracy? For one thing, the average age of exit pollsters has risen from 34 to 42, in response to the theory that they might unconsciously over-approach their generational peers. Edison-Mitofsky has also undertaken more training for exit pollsters on how to maintain a good random sample, and wherever possible, arrangements have been made with election officials to secure better and closer physical locations for the distribution of questionnaires.
In terms of what we will “know” when, the most dramatic change–first undertaken in 2006–will be the isolation of exit poll compilers and analysts to reduce leaks. The networks who are paying for the whole show won’t get access to any data until 5:00 p.m. Eastern Time, when two “waves” of exit polls are in. Once actual votes begin to come in, of course, the exits will be “adjusted” almost continuously, so you can expect the political commentariat to spend early Election Night trying to figure out what the prelimary analysis really means, and who is saying what based on which “wave” of exit polls.
On the other hand, as Matt Yglesias has already pointed out, a number of battleground states–FL, NC, VA, IN, and OH–that John McCain must win will be among the first to close the polls. If Obama’s up in most or all of them in the early exits, word will get around, with a hard-to-define impact on turnout in the vast majority of states where polls are still open.
The really crazy juncture will be right around the time raw votes start tumbling in, and the analysts try to figure out if the exits are on or off this year. Back in 2002, you may remember, the whole exit poll system crashed, leaving Americans with the archaic experience of having to wait on actual returns. That’s unlikely this time around, but you never know til Election Day arrives.