Larry Sabato today makes an excellent point about the Georgia Senate race and the national fight for 60 seats . . .
And we hate to bring this up, but Georgia has a requirement that the winner receive 50% plus one. The contest between Sen. Saxby Chambliss (R) and Jim Martin (D) includes a Libertarian candidate receiving up to 8 percent of the vote in polls. If neither Chambliss nor Martin can cross the 50% threshold, there will be a run-off election between them on Dec. 2nd. Could the battle for the 60th Democratic Senate seat extend the length of the eternal Campaign '08?
So is this a realistic outcome? What are our real chances here, and how do we win? Let's look at the polls . . .
First, where do we stand now?
The Pollster.com composite shows a race that is neck and neck, perhaps barely leaning towards Chambliss to the slightest degree. Note that this composite comes from a broad range of surveys that use a number of different turnout models. Many of these surveys likely use an older or outdated turnout model that is likely to favor the Republican incumbent.
To put it mildly, no one has the faintest idea what voter demographics are going to look like this year in Georgia, but we can start with a model of the past election and work from there.
In 2004, 25% of the electorate in Georgia was African American, 4% Latino, 1% Other, and 70% White. In that year's Senate Election, Johnny Isackson defeated a very poor Democratic Challenger (Denise Majette) by a margin of 58% to 40%, a margin that was essentially identical to the Bush margin of victory. We can pretty much use those values as a baseline of the best a Republican can do and the worst a Democrat can do in a statewide race that occurs during a presidential year. Our question then is what needs to happen to achieve two possible outcomes.
- Bring the Republican below 50%
- Bring the Democrat above 50%
Note that these two options are not the inverse of each other. Because Georgia law requires a run off if no candidate achieves 50%, we could very easily achieve a situation where the election is not determined on November 4th.
Here is the breakdown of the Senate vote from 2004, courtesy of CNN exit polling data:
(click image to see at bigger size)
By Algebraically manipulating these cells (that's a fancy way of moving these around), we can come up with a few different pathways towards acheiving our goals.
1. Increasing the African American vote, and increasing the proportion of it that votes for the Democratic candidate. This is clearly a major push for the Obama and Martin campaign, and one that is likely to be highly successful relative to 2004 in that a greater proportion of the Georgia electorate is likely to be AA this year than in 2004. A key point though, this tactic in and of itself is unlikely to be sufficient to win either the Senate or Presidential race. Current early voting returns show the AA vote percentage at 36% of currently cast ballots. This fantastic showing (if combined with an increase in the proportion of AA's that vote Democratic to 97% would be enough to swing the race. However, when this race is finished it is highly unlikely that AA's (who make up only 29.9% of the population) will be able to over perform their demographic by this margin. If they do, fantastic, but its not something we can count on.
For argument's sake, lets assume that AA's improve their share of the electorate to 30%, to equal their percentage of the population and then let's assume they break 95 to 5 for the Democrat.
Those Demographics would leave us nearly 10% points behind Chambliss, and him still over 50% achieving neither of our goals. Clearly we need to look in other places.
2. Flipping the Latino vote. This is an unlikely avenue for us also. In the 2004 election, Latinos broke for the Republican 53% to 46%. For the sake of argument, and my anecdotal information from making phone calls supports this, lets suppose we flipped this vote exactly so that it instead favored the democrats. This would only slice 1% off the lead outlined above leaving us 8 points behind and Chambliss still over 50%.
It seems to win this thing, the road is going to have to travel through Whites.
3. Win more White votes for Jim Martin. In 2004, the combination of a Kerry lead at the top of the ticket and an AA Female (who was also an awful candidate) conspired to produce record low percentages for the Democratic Senator and Presidential candidates. The big question out there is how winnable is this Demographic?
The answer is, its in play to the extent that could win us the election. keep in mind that if we achieve the goals outlined in 1 and 2, we don't need to come anywhere close to winning whites to win the election for Martin (and perhaps by extension Obama). For example, if Martin were the win the 35% of whites that Gore did in 2000, that would propel Martin to a 53.5/45.5 victory given the conditions listed above.
Still, electoral conditions in Georgia including the concentration of hard core Republicans in the North Atlanta Suburbs, make reaching 35% a stretch. What would be the minimum percentage of whites that we could win and still have a feasible shot of either of our two objectives.
Some have argued that we would need to stretch all the way up to 30%. This is true to achieve a victory outright, but not true to achieve a runoff.
4. Work for a run off by heightening libertarian ideology. This last strategy is an interesting one, and it assumes we could make short work of Chambliss in a post-wave December run off election - this is certainly a less desirable option than winning outright, but is also an infinitely better option than losing outright.
SUSA currently has Libertarian candidate Alan Buckley running at 5%. If we assume that 2/3 of the Buckley vote comes from people who would vote Republican otherwise and 1/3 from Democrats and subtract these from a baseline election result where Chambliss wins 75% of the white vote and Martin wins 25% (a very easily obtainable result), and we assume roughly 6% of the independent votes are coming from Whites, and we split the reductions 2 to 1 from the Republicans, when we keep the conditions above intact then we end up with a 48.9% to 45.0% Chambliss margin and a run off election (assuming votes from other races split evenly).
Here's the take home message: We force this election into a run off even when our share of the white vote is as low as it was in 2004 (22%)!!!!
Based on the above, here is a sort of fantasy turnout and voting table for the 2008 Senate elections:
After doing this analysis, I am more convinced than ever that the Georgia Senate seat (and perhaps the Georgia Presidential) are hanging by a thread if that for the Republicans. This is an eminently winnable race and comes with the added consolation prize of forcing a run off if things don't meet up to our expectations.
What are the keys?
- An African American vote percentage of at least 29%, roughly proportional (a little behind actually) to their current share of state population. Getting our share of that vote up to 95%.
- Ceding no ground on the Latino vote.
- Maintaining or exceeding the same share of the White vote that Denise Majette received in 2004 (again, this is likely the floor of the lowest the white vote could go. Al Gore won as much as 35% of this demographic as recently as 2000).
- Ensuring a 3% to 5% showing for the Libertarian candidate.
If these conditions can be attained give or take a point here or there. We will almost certainly force a run off for Senate. Any deviation upward will give us the win.
Conclusion: This is a race worthy of fairly massive financial investment over the last 12 days of the campaign.
[Update 2:17 pm] From Ambinder:
So African Americans make up 36% of the early vote turnout in Georgia.
A Republican who is tracking the data predicts that black voters will make up 35-36% of the electorate -- C + 5.
That is -- their proportion in the census (29%) plus five points.
An Obama campaign aide tells me separately that many of these voters are sporadic presidential voters.
And in 2004, black voters made up about 25% of total turnout.
Doesn't mean that Obama's going to win Georgia...
It just means that he'll do a lot better there than John Kerry.
From this its clear that most columnists simply don't understand the math of this race (as I did not this morning before this exercise)
Marc Ambinder is very smart and astute, but if AA's really end up making 35 to 36 % of the electorate than both an Obama and a Martin win are assured.