November 05, 2004

The Skinny on the Exit Polls, or how academia ruined the polling data

By now all of you have heard that the biggest fluster-muck of the MSM's coverage of the election was the pathetic accuracy of the exit polls. Since I have some first hand information on how the process worked, at least in this part of my little state, let me share some quick thoughts.

Some in the blogosphere are calling this a conspiracy of sorts. That the numbers were deliberately cooked. Others say the numbers weren't deliberately pro-Kerry, but that their early release was part of a plan to depress Republican voter turnout in Western states. For instance, over at Powerline:

It seems likely that Democrats ran the exit polling process, and deliberately generated bad data to create momentum for the Kerry campaign. This much, I think is a reasonable inference. It is also reasonable to suspect that the same people who created the bad data leaked it to Democratic bloggers as part of a strategy of depressing Republican turnout.
David Limbaugh adds:
Personally, I don't know if Democrats ran the exit polling process, but I'm sure liberals did, which is essentially the same thing. And there is very little doubt that those in control leaked it....

Someone somewhere in the process deliberately manipulated either the polling process, or the selection of participants, or something. With the pre-election polls as accurate as they were, it's virtually inconceivable that the superior exit polls could have been off like they were. Something smells very rotten.

David is right, and I have some evidence to back that up.

As most of you know, I blog from my desk at a small university somewhere in the vast hinterland of America. From my earlier posts on the subject, you know that my university is just as liberal as Berkeley or any other campus (in my department 9 full-time faculty voted Kerry, 2 for Bush, 1 did not vote, and 1 declined to state). So, why is this relevant? Because all those who ran the exit polls were students and the students were recruited by faculty members.

Two weeks or so before the election I received a phone call from a colleague on the English faculty. She had been contacted by one of the firms that had been subcontracted to conduct the exit polling for the Associated Press. They wanted her to recruit three students to do exit polling in the local vicinity and as many students as she could get to travel to nearby counties. Why is this relevant? Because she is a junior English instructor. If the firm had simply looked up our university in the phone book I would think they would contact me--the senior member of the political science faculty and the specialist in American government. They did not. So why did they contact her? No idea. However, I will note that she is an activist in the Democratic Party, an open Kerry supporter, and involved in many liberal causes.

Later, I learned the firm contacted one other faculty member to recruit students--the director of Social Work. If you do not know what Social Work students are like think of a typical lefty student stereotype, add a dash of love for all things Great Society, and throw in more than a spoonful of feelings of entitlement and there you go. The firm recruited all of their workers to conduct polling from among this group.

All of the polling in 1/8 of our state was done by Social Work students. All of it. My state, if you must know, was said to be 'too close too call' according to the exit-polling data. In actuality, Bush won by 10 points.

If the exit polls were conducted by students all across the country, I suspect that we may have one of the reasons why the initial data turned out to be so wrong. Students, generally, are more liberal in college than when they get out into the real world. Further, if the type of student recruited at my university was repeated across the country then we have an even bigger problem. The firm contacted two very liberal professors (even more liberal than the rest), who in turn asked their favorite students if they wanted a chance to make some cash.

From there it is not hard to figure out why the numbers would be so skewed, even if the students did not deliberately do it. You see, each of us has a natural tendency to feel more comfortable talking with some type of people over others. If I was to conduct an exit poll, I would feel quite comfortable talking to a guy in a cowboy hat. On the other hand, a guy wearing a Che shirt and Birks might be somewhat out of my comfort zone. Not that I wouldn't talk to any Che shirt guys, just not as many. If my bias is repeated over and over, by hundreds and by thousands of poll workers, than this bias becomes substantial when the numbers are aggregated.

This explains why the data were wrong. Students, possibly a lot of very liberal students, were hired for a day to run the exit polls. They tended to oversample other liberals--not in a huge way but enough to make a difference. These data were then turned over to the experts who aggregated them and then reported bad data to the major news outlets.

Two questions remain. First, why was it that a well known liberal and the head of a department renound for indoctrinating liberalism into it's students were contacted to recruit students? It may just be a coincidence, but I fail to see the logic in it. Second, why were the misleading data leaked out? Our explanation, thus far, only suggests why the data were bad. It does not explain why these data were leaked out. There may be no connection. It only takes one excited Kerry supported to call Wonkette and Sullivan and then let the 'good news' pick up steam from there.

Just a few thoughts. Many questions remain.

By Rusty Shackleford, Ph.D. at 01:19 PM | Comments |