wpostServer: http://css.washingtonpost.com/wpost

The Post Most: Politics

Read In

Now Viewing: People from around the country looking at Post Politics section

See what's being read across the country ›

Social Surface: Politics

Behind the Numbers
Posted at 12:13 PM ET, 11/04/2008

Where are my exits?

Here's a brief overview of what to expect today from the exit polls and some tips about how not to be misled.

First, you will need to wait a bit.

The networks were not appreciative four years ago when early exit polls were leaked and widely misinterpreted, spurring premature Democratic glee and some to congratulate "President" Kerry. Following that debacle, the National Election Pool, the consortium of networks and The Associated Press that run the exit poll, set up a hermetically-sealed quarantine room for use on Election Day. Only three representatives from each of the sponsoring organizations are allowed in - they are denied cellphones and Internet access and they are also escorted to the bathroom.

Instead of getting data at 1 p.m. as in previous years, no one other than those in quarantine has access until 5 p.m. This arrangement held in 2006 and throughout this year's primaries. There is no reason to think it will break down today. Therefore, any numbers you may see before 5 p.m., even those that purport to come from the exit poll, are likely fabricated.

At 5 p.m. others at the networks and subscribers, including The Washington Post, get access to certain exit poll data. We have subscribed to the national exit poll and state polls essential to our election coverage.

Some will post numbers shortly after 5 p.m.; we will wait a bit to assess the validity of the data. No matter what you see, please remember that these early numbers are very preliminary - only interviews conducted through early afternoon will be included in the first release; in the case of west coast data, it'll be only morning interviews. Moreover, the first round of data has been adjusted to expectations gleaned from pre-election polling and historical vote patterns. These may be off the mark in this epic election, or not.

Whatever the case, these data should be treated with great caution. As the night goes on, we will get more and more complete information from the exit poll and from actual vote returns. It is a process.

Even as the data improve throughout the night, however, one should never forget that the exit polls do not indicate who is going to win. They are simply not designed to provide predictions.

The AP, the television networks and Edison-Mitofsky, the firm that conducts the poll, use exit polls as one element of complex election-calling decision models. Exit polls are valuable primarily for the estimates of who voted and why, with the most useful data being the crosstabs (e.g., percent of women supporting Obama, independents McCain) and estimates of voter characteristics (e.g., percent African American, white evangelical).

You may see some early estimates of the vote by state, or think you can tease out of network coverage the answer to the basic "who is winning?" question, but do not be misled. Early exit polls are often out of line with final results - that is not because they are bad polls but because they are not magically predictive.

We will highlight some of the significant findings here at Behind the Numbers, but as you dig through whatever data you get a hold of, please remember exit polls have error margins, just as all polls do. One relatively safe rule-of-thumb is to look only at double-digit advantages as significant.

One note on early voters: The national "exit" poll includes a parallel telephone sample of those who cast their ballots before today. Same is true in 18 states of greatest interest where early voting has been historically high. These telephone polls do not include parallel samples of cellphones.

Another note on reports of a Democratic bias in recent exit polls: In order to minimize any such skew in the 2008 exit poll, Edison-Mitofsky improved its interviewer training program, emphasizing how essential it is to stick to the sampling plan. They also aimed to hire a more diverse base of exit pollsters that has raised the average age of interviewers by nearly 10 years over the 2004 level (34 years old).

The exit pollsters have also taken a more active role, ultimately successful, in contesting state laws attempting to keep interviewers too far from the polling place to reach a good sample.

Here are some links to help you bide time until the exits come out:

Pew's Andrew Kohut's interview with Edison-Mitofsky's own Joe Lenski about this year's exits.

StarTrib story about the exit poll lawsuit in Minnesota.

WaPo story about the 2004 exit poll evaluation report.

Pollster.com overview on the nuts-and-bolts of exit polls.

By  |  12:13 PM ET, 11/04/2008

Categories:  Exit polls

 
Read what others are saying
     

    © 2011 The Washington Post Company