Wednesday, August 20, 2008

Confidence in numbers

Come November 4, the electorate will be markedly different from the 2004 composition of evangelicals, white males and national security-concerned voters that held the White House for George W. Bush.

This is particularly important when one considers the narrowness of the Republican's victories over the Democrats in the key battlegrounds of Virginia, Colorado and Ohio. In Virginia, McCain starts with an inherited Bush majority of 262,217 votes, in Colorado 99,523 votes, and in Ohio with 118,599 votes. Thus, it is immediately apparent that a collection of small changes in the make-up of the electorates of these key battlegrounds can have a decisive effect upon the outcome.

Compared to 2004 and given the scale of Obama's ground game in terms of voter registration and, later, Get Out The Vote operations, his performance amongst key demographics in the primaries and the changed political climate as a whole the following general predictions as to the 2008 electorate's composition are now offered:

- Under 30's: up
- Over 65's: down*
- Hispanics: up
- African American: up
- Evangelicals: down
- National security as No. 1 issue: down
- Economy as No. 1 issue: way up
- Cultural issues: way down

Each of these electorate changes is to the benefit of Obama at a cost to McCain.

Furthermore, current polling, especially those that utilise a Likely Voter screening model as opposed to a Registered Voter polling model, generally tend to under-sample these categories (for example, this poll which shows under 30's as just 6% of the electorate instead of the 17% they were in 2004) and as a whole thus skew Republican. Because these target groups that are particularly susceptible to voter registration and GOTV efforts they are likely to be underrepresented in pre-election polling as opposed to their actual Election Day strength.

Sound familiar? That's because it is. This is the same model the Obama campaign employed to such great effect in Iowa. As such, the majority of polls failed to call the race accurately, as they predicted that the Iowa caucus-goers composition would be broadly similar to 2004 whereas in truth it was transformed by the new caucus-goer registration efforts and the subsequent GOTV operation.

Herein lies the importance of the unprecedented scale of Obama's field operation. This election may well prove to be Iowa writ large.

* = whilst the absolute number of over 65 year old voters may be up, as a percentage of the electorate as a whole it is likely to be down.

1 comment:

John said...

Excellent piece. I would certainly agree that, both the undersampling in the polls and the ground effort by the Obama campaign logically suggests that Obama will 'overperform' in November, but can we find some evidence about large this will be, maybe by looking at the evangelical turnout for the republicans in 2000 and 2004?