Georgia Senate Runoff Analysis

Why did Jim Martin lose so badly in the runoff when he had done so well in the general election?

This is the last in my series analyzing the 2008 election in the state of Georgia. I looked at statewide numbers in my first article. The next article looked at county level results and also racial polarization. In the 3rd article, I analyzed the results in Fulton county and specifically the precincts in which our team of Obama volunteers worked.

Georgia has a law according to which a winning candidate needs 50% + 1 of the votes cast, otherwise there is a runoff election between the top two candidates. In the general election on November 4, incumbent Republican Senator Saxby Chambliss failed to get 50% and thus was forced into a runoff with the Democratic candidate Jim Martin. Interestingly, this was the second runoff for Jim Martin this year since he had won in a runoff election in the Democratic primary too after no one got 50% in the crowded filed of Democratic Senate candidates in the primary election.

I did not work for the Jim Martin campaign and in fact was fairly certain that he would lose the runoff because Georgia is still a Republican state and there are a lot more habitual Republican voters here than Democratic ones. The Obama campaign had done a good job of registering voters and then bringing the sporadic voters out to vote for the general election, but there was no way that could be repeated for a runoff. Also, from our voter contact when we were volunteering for the Obama campaign, we knew that Sarah Palin was quite popular among Republicans here, perhaps more than John McCain (McCain got 31.6% of the vote and came in 2nd behind Huckabee and barely ahead of Romney in the Feb 5 primary.)

Chambliss Martin Total
Votes % Votes % Votes
General 1,867,090 49.75% 1,757,419 46.83% 3,752,577
Runoff 1,226,730 57.46% 908,222 42.54% 2,134,952

As you can see, a lot less people voted in the runoff than in the general election. Let’s look at the voter retention rate, i.e. the number of votes for a candidate in the runoff as a proportion of the votes in the general election.

Chambliss Martin Total
Vote Retention 65.7% 51.7% 56.9%

While only about 57% of the voters in the general voted in the runoff, they were not evenly distributed between the parties. Democrat Jim Martin barely managed to get half of his voters to pull the lever for him again while Republican Saxby Chambliss was successful with almost two-thirds of his voters. This is why the Republican margin increased from 2.92% on November 4 to 14.92% on December 2.

Let’s plot the percentage of votes the two candidates got in each county in the general and the runoff elections. The circles represent individual counties and their area is proportional to the number of registered voters in that county.

Chambliss percent general vs runoff

Martin percent general vs runoff

Chambliss performed better in all counties and Martin performed worse in almost all in the runoff when compared with the general election.

Another way to look at the data is to compare the vote retention rates for Chambliss and Martin.

Vote retention Chambliss vs Martin

Chambliss got about 60-70% of his general election vote in most counties, especially the more populous counties while Martin only got 45-60% in most. The only counties close to parity in terms of vote retention were really small.

I was wondering whether there was any pattern to the vote retention in terms of partisanship. So I plotted the Chambliss and Martin retention rates against the percentage in that county that voted for Obama on November 4.

Martin vote retention vs Obama vote

This is a slight upward trend as the counties get more Democratic. However it is still very low.

Chambliss vote retention vs Obama vote

We see much less of a pattern here.

Georgia Election Analysis III

Today I dig deeper into the Presidential election results in Fulton county and specifically the precincts in Alpharetta, Milton and South Atlanta where our team of Obama volunteered worked.

I did some analysis of the Presidential election in the state of Georgia in two installments. Today I’ll look at Fulton county where I live as well as the specific precincts in Alpharetta, Milton and South Atlanta where our team worked as volunteers for the Obama campaign.

Year D R Turnout %Turnout
2000 57.53% 39.68% 264,276 65.06%
2004 59.35% 39.99% 330,791 74.15%
2008 67.07% 32.09% 405,531 73.11%

There was a 15.3% swing towards Obama in Fulton county compared to Kerry in 2004. The turnout, based on the number of registered voters, stayed constant from 2004 to 2008. However, if we use the voting age population estimates (VAP) for Fulton county, we get the following turnout rates:

Year Turnout based on VAP
2000 42.30%
2004 48.00%
2008 52.72%

This shows a bigger turnout in 2008. One reason for the discrepancy between the turnout in the first table and this one is that the Obama campaign focused a lot on voter registration this year and thus got more people registered.

Also, please note that the VAP turnout estimate is probably lower than the real turnout which should be calculated as a proportion of the voting eligible population.

Out of the 3,924,440 votes cast for President in Georgia this year, 2,084,179 (or 53.11%) were cast during advance/early voting or by absentee mail-in. In Fulton county, 184,240 votes (45.42%) were cast early, absentee or provisional out of a total of 405,628 votes cast. This is very unusual for Georgia and Fulton as can be seen in the table below for Fulton county. (This shows the numbers for the 2004 general election for President, the 2006 election for Governor, the 2008 Presidential Primary on Super Tuesday and the 2008 general Presidential election.)

Year Absentee/Early/Provisional votes
2004 Prez 12.68%
2006 Gov 11.76%
2008 Prez Primary 7.65%
2008 Prez 45.42%

In the 2008 general election, the Obama campaign tried to get everyone to vote early. The effect of this can be seen in the early voting numbers in Fulton county where Obama had a 49.5% lead in early voting compared to a 22.9% lead on polling day itself.

The large numbers of early voters have complicated my precinct level analysis. The precinct level data does not include early or absentee votes which are listed separately as one per county. Still let’s see what we can conclude for our precincts.

For the Get Out the Vote (GOTV) campaign in October and then the first four days of November, we were based in the South Atlanta precincts, 12E1, 12J and 12T, which are just north of Atlanta’s Hartsfield-Jackson Airport. These precincts are heavily Democratic (more than 90%) but there are a lot of sporadic voters there. So the main task was to get the vote out.

Election 12E1 12J 12T
Voted %Turnout Voted %Turnout Voted %Turnout
2004 518 49.5% 1024 59.7% 566 62.3%
2006 224 22.1% 473 29.2% 581 35.8%
2008 Primary 280 27.5% 564 34.6% 609 39.1%
2008 747 55.1% 1,360 67.3% 1,232 68.0%

Thus, turnout as a percentage of registered voters increased from 57.4% to 64.3% while actual number of people who voted in these three precincts increased from 2,108 to 3,339, a 58% increase in the number of voters. Compare this 58% increase to a 22.6% increase over the whole of Fulton county and a 19.6% increase over all of Georgia. I am actually pleased at these numbers and I am sure all of our Obama team would be too.

Let’s now look at the precincts in Alpharetta and Milton where we made the most calls, starting in August, for identifying voters, persuading them and then getting them out to vote. The table below shows the Republican advantage over Democrats. For the 2008 Presidential Primary, I added the votes for all the Republican Presidential candidates together and did the same for the Democrats. Let’s take the example of ML01A in the 2008 general election, McCain got 73.5% and Obama got 25.3%, which means a Republican advantage of 73.5-25.3=48.2%.

Precinct 2006 2008 Primary 2008
AP07B +66.0% +33.2% +40.9%
ML01A +64.0% +34.4% +48.2%
ML02A +67.4% +40.8% +62.6%
ML02B +68.0% +45.2% +57.5%
ML03 +66.9% +45.7% +59.5%
ML04 +65.2% +38.5% +56.5%
ML05 +48.9% +14.9% +37.7%
ML06 +29.9% -11.4% +8.2%
ML07 +65.1% +41.8% +52.0%

I couldn’t find the data for the 2004 Presidential election for the Milton (MLxx) precincts but AP07B had a 42.9% Republican advantage in 2004.

Before I opine on those results, let’s look at the turnout (as a percentage of registered voters) in these precincts:

Precinct 2008 Turnout
AP07B 44.6%
ML01A 44.6%
ML02A 55.3%
ML02B 55.3%
ML03 52.5%
ML04 44.7%
ML05 44.2%
ML06 41.9%
ML07 54.2%

We already know that about half of Georgia and Fulton county voters voted early. The early voters are not listed in the last two tables for the Alpharetta/Milton precincts because they are not reported that way by Fulton county. Looking at the election day turnout for Milton, it seems that fewer than half of the voters voted early.

Because we only have the vote breakdown by party for those who voted in person at their polling location on November 4 and a significant number of voters had voted early in 2008, we cannot really say what the margin between McCain and Obama was in Alpharetta and Milton. There is no reason to believe that the early voters had the same partisan distribution as those on election day. In fact, there is reason to believe that the early voters were more likely to be Obama supporters. The Obama campaign had been working hard asking people to vote early. Also, in the whole of Fulton county, Obama had a 49.5% advantage in early voting compared to a 22.9% advantage on polling day. Thus, the very large Republican margin in the Milton precincts in 2008 is most likely not correct. I can say with certainty that McCain won all those precincts except ML06 but probably with a somewhat smaller margin. Even then, Republicans had a huge advantage here in North Fulton. Unfortunately, we don’t have any way of finding out whether our team was able to reduce that Republican advantage or not.

Next: A look at the Senate runoff between Jim Martin and Saxby Chambliss.

Georgia Election Analysis II

I look at the detailed county level results for the 2008 Presidential election and also try to quantify the effects of race.

Continuing with my election analysis in Georgia, my data source is the Secretary of State’s website:

Let’s start with a graph showing Obama’s share of the vote in each Georgia county versus Kerry’s share four years ago. Please note that the size of the circles is proportional to the total number of registered voters in that county.

Obama vote vs Kerry vote

Since there was an 11.4% swing towards Obama compared to Kerry, a majority of the counties show a larger percentage for Obama as compared to Kerry. This is especially true of large (in population) counties. Since Georgia has 159 counties, let’s redo the same graph separately for the large, middle and small counties.

In the largest 53 counties (min number of registered voters 17,354 and max 554,682), only the most Republican counties got more Republican compared to 2004.
Largest counties swing

In the middle 53 counties (registered voters ranging from 8,182 to 17,018), the situation was a lot more mixed, but Obama did lose a lot of ground in some.
Swing in middle counties

In the 53 counties with the lowest populations (registered voters ranging from 1,371 to 7,988), Obama did better in some and worse in others. However, these rural counties don’t have many people and probably didn’t matter much at all for the Presidential election.
Swing in least populated counties

Since one major reason for Obama’s better performance was the higher turnout and vote from African Americans, let’s look at the same data, i.e. Obama’s total vote share compared to Kerry’s, but by categorizing counties by how white they are. I am using the percentage of registered white voters in a county as a proxy for the percent population that’s non-Hispanic White in a county.

The whitest third of the counties are at least 77% white. Do note that the scales of this graph are different from the earlier graphs. Obama did worse in the smaller, whiter counties than Kerry and better in the bigger white counties. But he didn’t get more than about 35% of the vote in any of them.
Swing in the whitest counties

The middle third counties by white percentage of the population have about 64% to 77% whites. Obama generally did better than Kerry here. Obama’s share of the vote varied from 27-45% in these counties except for Clarke county (65%) which contains Athens and the University of Georgia.
Swing in the mid-white counties

In the bottom third counties, the white percentage is 18-64% and Obama did very well here.
Swing in least white counties

Let’s look at Obama’s share of the vote in each county plotted against the percentage of whites and blacks in that county.

White percentage vs Obama vote share

Black percentage vs Obama vote share

The red line in both graphs is the least squares fit. As is obvious, Obama’s share of the vote decreases as the percentage of whites increases in a county.

In the previous two graphs, we looked at Obama’s share of the total vote. So we cannot completely tease out the effects of a large African American turnout or the how the white voters voted in each county. Since we can’t have the numbers for Obama’s share of the white vote in each county, we’ll make some simplifying assumptions to guesstimate it:

  1. The African American turnout in each county was the same as the total voter turnout in that county. This is approximately true at the state level, but could be wrong for some counties.
  2. 98% of the African American voters voted for Obama. This is the percentage at the state level.
  3. The number of voters who are not African American or White is small enough (5%) that we can approximate the white vote share by simply excluding the African American vote.

We know this procedure gives us only a guesstimate of the nonblack vote since in one small county (Clay) this results in Obama getting -1.85% of the nonblack vote. In the other 158 counties, Obama’s share of the nonblack vote varies from 4.4% (Randolph county) to 54.6% (Dekalb county).

White population percentage vs Obama share of nonblack vote

Black population percentage vs Obama share of nonblack vote

The red lines in both graphs are LOESS fits.

The first thing apparent in these two graphs is that Obama did particularly well among whites in most of the populous counties as they show up outliers at the top. Among the other counties, there is not a significant pattern, but it does look like Obama’s nonblack vote share is lowest in counties where African American population is in the 20-30% range.

Razib did a similar analysis of racial polarization in Mississippi.

Charles Franklin analyzed the white vote for Obama at the state level. Nate Silver saw that data and noticed that most of the states where Obama performed worse than Kerry among whites were those where Obama did not campaign.

Andrew Gellman, at his must-read blog Red State Blue State Rich State Poor State, looked at Obama’s vote share divided by race in the West, Northeast, South and Midwest. He also showed the estimated nonblack vote by county for all of the US.

Andrew Gellman also looked at the voting preferences of rich and poor voters in both red and blue states in the 2008 election, something which is the topic of his book Red State, Blue State, Rich State, Poor State: Why Americans Vote the Way They Do.

Finally, Michael McDonald estimates the voter turnout from the vote eligible population.

My revised national turnout rate for those eligible to vote is 61.6% or 131.3 million ballots cast for president. This number does not include an approximate 1.4 million under and over votes, for an estimated total 132.7 million total ballots cast. This represents an increase of 1.5 percentage points over the 60.1% turnout rate of 2004, but falls short of the 1968 turnout rate of 62.5%.

His estimate of turnout for Georgia is 61.1%, up from 56.2% in 2004.

Next: An analysis of the precincts in Alpharetta, Milton and Atlanta where our team worked.