Saturday, October 18, 2008

2008 Ohio Vote Projections: McCain's Margin of Error Less Than 1%

UPDATE: Tonight I'll be comparing this projection against the actual votes to determine how each candidate it doing to predict Ohio early.

I've completed my initial projections of Ohio's vote for the 2008 presidential election based on the current registered voter list and past behavior. The bottom line is McCain has some room for deterioration from Bush's 2004 performance. But not much.

I did my preliminary projections at the county level. Hopefully I will be able perform the same analysis at a precinct level to see if there is much of a difference.

To do my preliminary projections, I took the 2008 voter registration counts for each county, applied a turnout factor equal to the turnout for the county in 2004, and then applied the a factor representing the percent of votes Bush got to determine a McCain baseline and a factor representing the percent of votes Kerry got to determine a Obama baseline. I did this for each county and totaled them up for a statewide total.

Using this method, McCain wins Ohio by 125,928 votes. This was a slight improvement over Bush's performance (118,000 votes) because Bush counties actually showed a slightly larger increase in registered voters in 2008 than Kerry counties.

That means, if everything mirrors 2004 (turnout, margin of victory in each county), McCain wins Ohio by a little over 125,000 votes.

But everything remaining the same is highly unlikely, and the things that will change favor Obama. The question is, how much will they change and how much will they eat into the 125,000 votes?

The first factor to consider is turnout. Turnout in Bush counties was 72.29% while turnout in Kerry counties was 68.86%. After the 2004 election, Ohio's voter lists were purged of invalid voters with a significant number coming from the urban counties that favored Kerry. These invalid voters artificially lowered the turnout in Kerry counties to some extent and you would expect higher turnout in 2008. What's unclear is how many of the new voters registered are invalid. It's clear that ACORN has been registering some invalid voters but it's not clear how many really made the actual registration list, etc.

For the purposes of this analysis, I'm going to assume the turnout in the Kerry counties increases to the same level as the Bush counties in 2004. This would require an 5% increase in the number of voters turning out in Kerry counties.

This increase in turnout would net Obama an additional 24,572 votes.

This means that even if Obama raises the turnout in Kerry counties 5%, matching the turnout percentages in the Bush counties, he will still lose Ohio by over 100,000 votes.

Obama will not win Ohio by just increasing turnout in his base counties. He must also eat into Bush's margin of victory.

There are a couple ways he could do this. Bush won his counties in 2004 by an average margin of victory of 20.90%. Kerry won his counties by an average of 16.62%.

If Obama can win the same Kerry counties by 20.90% like Bush won his, he brings the election to within 2,500 votes. Factor in the increased turnout in the Kerry counties and he wins by a little over 20,000.

Likewise, if Obama can reduce McCain's margin of victory in the Bush counties to 16.62%, he brings the election to within 5oo votes. Again, factor in the increased turnout in Kerry counties and he wins by 24,000.

Another option. Increase the turnout in the Kerry counties as stated above and increase his margin of victory by 0.9% in Kerry counties while reducing McCain's by 0.9% in Bush counties.

The result: a 5,849 vote victory for Obama. Is that close enough for you?

In a future post, I would like to perform a similar analysis at the precinct level. Using a county rollup could hide trends in voter registration that I am missing.

I would also like to look at trends in 2000 to 2004 results and see what happens if I project them out into 2008.

The bottom line is something everyone knew. McCain has very little margin for error and can't allow even a 1% deterioration from Bush's margins of victory if he is to win Ohio.