Gyroscope

A newsletter for those unmoved by spin.
No. 51, November 1, 2004

Subscribe

Change Address

Unsubscribe

Comment

You won't get much help from the press in the early hours to decide how things are going. They have their canned story lines to follow and few of the reporters will have spent any time assessing the data or know how to do a useful analysis.

However, we are here at Gyroscope to solve your dilemma. At 4:30 pm PST, 11 states will have closed their polls. By comparing the actual results against what polls in those states have been predicting, you can assess if their is any trend.

The complication is that we won't know at 4:30 the final results for all vote counting for those states, what we will know is if the networks have called the state for one candidate or the other. But we can, based on polling data, decide what we expect the networks to do. If the race is lopsided in a state, the networks call it for the candidate almost immediately after the polls close. If the race is close, they only call it some time after the polls close.

Here is how it works. Taking state by state polls across the country winds up with a narrow electoral vote victory for Bush. I'm aware that a number of recent polls are showing a slight break for Kerry, but that hasn't translated into state by state polls as yet. So here is what we have

So, at 4:30, if the race is close, Bush will have 6 states, Kerry only one. This will look like "an early lead for Bush" but it's not a fair cross section of the nation.

The key for our analysis is to examine changes from the expected result. Some states are so far to one side or the other that even a significant change from expectations won't be noticeable. If Bush is going to win Indiana by 10 instead of 19, that is pretty significant, but it would still cause the networks to make an early call for Bush and so we won't notice anything different.

The test comes with the four states where the result is in bold face. Those are places were we will first see it if the election is going to go differently than the narrow Bush victory the polls predict. If at 4:30, the networks call Florida, New Hampshire or Ohio for one candidate or the other, that is going to be an early indication of a trend. If Virginia stays undecided into the evening, that would be a sign of a Kerry trend.

If Kerry were to loose Florida and Ohio, he has a hard time winning the election. If he wins them both, he becomes the favorite to win the rest of the nation.

This analysis assumes that the networks' calling of the states is done on purely statistical grounds and that they are being careful. They weren't careful in Florida in 2000, and some analysis I've read suggest that they are even more dependent on a single source of data then they were in 2000. And the pressure to call it first is so strong that ideology may trump statistics once again.

Also, I'm assuming that trends are national. If Kerry does better than expected on these east coast states, then I assume he will do better across the nation. That's not a bad assumption, actually, but still an assumption.

There is another problem to consider with our analysis. The state by state polls predict a slight win for Bush. So if Kerry is doing a little better than expected, that means the race is tight, not that he will win. He'd have to do significantly better than expectations to be put slightly in the lead.

I also don't know what the criterion is for calling a state, so I don't know if a 5 point eventual victory is enough for them to call a state immediately or make them wait.

However, I think this is a pretty good set of indicators, and better than some the media uses. Media, especially local media, regularly reports that "the number of voters is higher than the previous election" evidently forgetting that we have a growing population. There are 217 million people over 18 in the U.S., up from 202 million in 2000, so even if turnout is the same percentage as last time, turnout will be a record. This matters for predicting the election because high turnout traditionally favors the Democrats and one thing that Kerry supporters are hoping for this time is a high turnout. Early evidence of high turnout should be carefully reviewed before us Kerry voters get too excited.

So what if 4:30 comes and we only see what we expect or a slight drift to Kerry? Well, at 5pm a whole batch of states close their polls. Here are the ones to watch:

Here we really have a good set of indicators to consider because we have an almost perfect symmetry with one state each by 5 for a candidate and one each by 2 or 3 for a candidate. If 4:30 hasn't revealed the answer, 5pm may do so. If at 5 one of the "not called" states has broken for a candidate or if one of the "most likely" states has not - then you've got evidence of a trend. Again, remember the overall narrow Bush lead we start with. So if Missouri is uncalled and New Jersey breaks for Kerry, it may mean the election is heading for a tie.

The news media will likely refer to some of the data I've built this analysis on, but they are - if this election is like previous ones - likely to garble the data or mix important information with irrelevant data.

But now you can predict the result yourself.