November 1, 2008

Another Caution: Ignore Projections Based Upon Exit Polling

Hatched by Dafydd

This is another tip for watching the returns Tuesday night: You cannot project the winner of a state from the exit polling, becuase by its very nature, the respondent pool is not representative of the voters themselves..

The respondent pool in any exit poll of a state comprises:

  • Those voters who waited to vote until election day and --
  • Who voted at one of the precincts where the pollsters were polling --
  • At the time of day during which they were polling --
  • Who volunteered to stand still and be asked a bunch of intrusive questions that, in spirit, violate the sanctity of the secret ballot and --
  • Who decided to tell the truth about how they cast their votes.

Needless to say (but try to stop me!), none of these characteristics fits the voting population as a whole. A very significant portion of voters in most states vote early or absentee; pollsters don't lurk at every polling place but only a select few; they don't stay there from opening to closing, but only as long as they have money to pay for workers; Democratic voters tend to be more eager than Republicans to talk to pollsters (they see kindred souls); and of course, the "PC effect" is at super strength when voters are being interviewed face to face, without even the anonymity of the telephone: Many people are apt to tell you that they voted the politically correct way, no matter how they actually voted; they don't want an interviewer to think them ignorant, bigoted, or unsophisticated.

Predicting how the state will turn out was never the reason for exit polling, because most good pollsters realize the two pools (poll respondents and voters) are quite divergent. Rather, it was supposed to be an informational tool for statisticians to study elections after the fact and determine why people voted the way they did.

How do they do that? It's actually interesting: Rather than using the poll to predict the actual vote -- setting the stage for losers to think "we wuz robbed" -- valid pollsters collect all the exit-poll data, then align the respondent pool to the actual vote in that precinct, rather than the other way 'round. This allows them to draw reasonable inferences about what caused voters to vote a certain way... "Voters for Barack H. Obama were most persuaded by arguments or policies A and B, while McCain voters were most persuaded by C, D, and E." This allows for a post-mortem on the entire election, giving a wealth of data and correlations to all parties.

But it's incompetence bordering on malfeasance to attempt to project from the unrepresentative exit polling to declare who will win the state. Unless, of course, the goal is to suppress the Republican vote by projecting an Obama landslide.

(Contrariwise, it's perfectly proper to project who will win a particular state based upon the partial count by precinct -- provided you know enough about each precinct and polling place to know which voting patterns correlate to victory and defeat. Michael Barone, author of the Almanac of American Politics, is particularly adept at doing this.)

So don' t be a sucker: Ignore any state projection based upon exit polling -- even if they're calling it for John S. McCain.

Hatched by Dafydd on this day, November 1, 2008, at the time of 6:28 PM

Trackback Pings

TrackBack URL for this hissing:

Listed below are links to weblogs that reference Another Caution: Ignore Projections Based Upon Exit Polling:

» Four Paths to Victory from Big Lizards
Looking at the current Real Clear Politics electoral map, it seems clear that John S. McCain has four possible paths to victory. Realistically, however, each begins with winning all the states that are currently considered toss-ups. It's still possible... [Read More]

Tracked on November 3, 2008 11:13 AM


The following hissed in response by: SlimGuy

There was a full scale review done of all the primary exit polls and it was determined that Obama was over rated by almost 7% beyond his actual support in the exit polls.

The followers of him were just like the PaulBots willing to go though the 10 minutes and pages of exit poll questions to tout their hero more than others which is what skewed the data.

I am a past professional at this and have access to data all the way down to precinct level for the last 15 or so elections and can tell which precincts are 'indicators' and such.

But then you still have to adjust that for changes in population/demographics and many other factors as the areas transition over the years.

The above hissed in response by: SlimGuy [TypeKey Profile Page] at November 1, 2008 7:41 PM

The following hissed in response by: SlimGuy

Yet another factor that will throw off the exit polls this year.

Each year more and more states allow early voting or are easing their requirements for absentee ballots.

The result is something like 7 or 8 new states this year came on board and now 38 states allow early voting.

The effect this has is to tend to remove higher educated and higher wealth people from the polling lines it has been shown.

Thus this also skews the data pool that the exit poll people have to choose from which may not reflect the population in general.

Also the exit polls from the primaries have influenced the polling models used by pollsters and many of the earliest states were not exposed to much of what was revealed later in the campaign and thus may not reflect the true mix of voter positions at the current time.

The above hissed in response by: SlimGuy [TypeKey Profile Page] at November 1, 2008 7:51 PM

The following hissed in response by: scrapiron

Exit polls are like the ones in 2004. The pollsters will call the election for the wrong person because people see no reason to tell them the truth. I would tell them the exact opposite of how I voted since it is none of their business. Same when they call, tell them a pile of crap and they eat it up.

The above hissed in response by: scrapiron [TypeKey Profile Page] at November 1, 2008 9:30 PM

Post a comment

Thanks for hissing in, . Now you can slither in with a comment, o wise. (sign out)

(If you haven't hissed a comment here before, you may need to be approved by the site owner before your comment will appear. Until then, it won't appear on the entry. Hang loose; don't shed your skin!)

Remember me unto the end of days?

© 2005-2009 by Dafydd ab Hugh - All Rights Reserved