Polling all over the place

This weekend, the political buzz was over the LAT poll that says Davis may survive the recall and that Bustemante would win the balloting in part II.

I am not surprised to hear that support for the recall would fall because the reality of it is getting closer. And given the huge registration edge the democrats have, I would expect Bustemante to do well. However, the problem is the LAT poll is so far from the other polls you have to wonder if something is fishy. The key will be to see if the scale of the trend they are seeing is picked up by any other polling service.

I still remember the Clinton-Dole race in 1996 and some of the polls had Clinton winning the popular vote anywere from ten percent to the high teens. Do you remember how many percentage points the final result was?

Eight.

Some big name pollsters had a landslide of epic proportions but they turned out to be wrong.

And of course, the worst case scenario of polling happened in Florida on election night 2000. How could the networks call the election for Gore? They used polling data that said he was going to win Florida easily. The whole mess was chronicled in Bill Sammon's book At Any Cost: How Al Gore Tried to Steal the Election.
I personally thought the title of the book was a little over the top because Gore didn't break any laws in Florida. He pursued legal remedies at every turn. And at every turn where an interpretation of law went against the Gore team, the Florida Supreme Court stepped in with dubious legal reasoning.

Sammon gave us the inside story of how the polling data got so messed up that night.

One problem is that polling doesn't pick up absentee ballots which Republicans tend to use more frequently so the size of Bush's vote in that group was underestimated which pollsters try to correct for. Also, Sammon mentioned a historic trend that exit polls tend to overstate the democrat vote. It is something that the pollster can't seem to figure out why it happens but it does and so they have to "estimate" the size of this slight distortion of the data in the sample size. Also, the voter news service sample size is actually quite small and they may not have choosen the best "selected precints" to do exit polling.

All of this to say, the polling environment in California could be unlike any other election. On one hand, it is an off-year so turnout could be very low. On the other hand, the "zoo-like" atmosphere with so many candidates may mess up turnout models that pollsters use to adjust their data. And finally, we simply don't know to what degree disappointment with Gov. Davis will translate into votes on part I or part II of the ballot.