Daily Polling Report 11/2
by CA Pol Junkie
A batch of Mason-Dixon polls are out with what appears to be good news for John McCain: narrow leads in Ohio, North Carolina, and Missouri and a 3 point Obama margin in Virginia. Since Mason-Dixon has a house effect of GOP +2.5 and there isn't any discernible movement overall, its results really aren't news. Survey USA is reporting that white voters in Virginia are moving toward McCain. That may be true, but they still have Obama ahead while undercounting both the support and turnout of African-Americans.
I think the turnout of African-Americans will be a major story on Wednesday morning. In the primaries, the pollsters were all caught off guard by the turnout and near unanimity of their support for Barack Obama. All indications from early voting are that we will see the same thing again on Tuesday, which will be pivotal in Virginia, North Carolina, and Georgia. An analysis of Georgia early voting and a projection of Tuesday night's results there is below the fold.
African-Americans are 29% of the population in Georgia, but in 2004
they were only 27.4% of registered voters and 25.4% of actual voters. Since then, there has been a concerted effort at increasing both those numbers so that African-Americans exert political power in proportion to their population. Step 1 was a big success:
Registered active voters, September 2008:
Total: 4,978,704 (100%)
White: 3,176,520 (63.8%)
African-American: 1,445,153 (29.0%)
Other: 357,031 (6.2%)
Step 2 is increasing the turnout rate of African-American voters. In early voting, the results were spectacular:
Total: 1,994,940 (100%)
White: 1,201,001 (60.2%)
African-American: 699,866 (35.1%)
Other: 94,073 (4.7%)
The African-Americans turnout rate in early voting was an astounding 28% higher than for white voters. The remaining electorate who hasn't already voted is as follows:
Total: 2,983,764 (100%)
White: 1,975,519 (66.2%)
African-American: 745,287 (25.0%)
Other: 262,958 (8.8%)
To calculate the final vote, we need to know how each ethnicity votes on average. Using the cross-tabulations of polls from Survey USA, Democracy Corps, Insider Advantage, and Research 2000, the average two-party white vote for Obama is 26.1%. This is a little higher than the 23% John Kerry received in the 2004 exit poll. Polling of African-Americans consistently undercounted Obama's support in the primaries, causing the "Reverse Bradley Effect" when their support ended up being nearly unanimous. Polls have shown up to 98% support for Obama in the general election, so I will use that as an estimate. Research 2000 reported Obama's Georgia support among Hispanics and other races to be 67-8. I don't believe Obama's support will be that high, so I will assume 75-25.
Next comes the turnout assumptions. In 2004, 80% of Georgia active registered voters cast their vote in the presidential election, including about 82% of white voters and 74% of African-Americans. We can assume that the whites will be equally motivated to vote for McCain as they were for George W. Bush so their turnout rate will be the same as four years ago.
In early voting, African-Americans turned out at a rate 28% higher than whites, but the state just tips to Obama if they turn out at the same rate as whites among election day voters. Here are Obama's victory margins if African-American turnout remains elevated relative to white voters:
0% (same turnout rate as white voters): Obama +0.04%
10% higher: Obama +1.26%
20% higher: Obama +2.46%
28% higher (same relative turnout rate as in early voting): Obama +3.40%
As long as Obama gets at least the mid-20's level support from white Georgians like the polls indicate, he has a good chance of winning.
In 2 days we make history! What did you all do today to Get Out The Vote?
The current prediction is Obama 378, McCain 160. Colorado is the state which puts Obama over 270 electoral votes while Indiana is the closest state, still very narrowly for McCain. The states really aren't shifting around relative to each other. They've all been polled to death and we have a consensus on where they fall on the Obama-McCain scale.
Obama Base (264 EV): California, Oregon, Washington, Hawaii, New Mexico, Iowa, Minnesota, Wisconsin, Illinois, Michigan, DC, Maryland, Delaware, Pennsylvania, New Jersey, New York, Connecticut, Rhode Island, Massachusetts, Vermont, New Hampshire, Maine
Competitive states, cumulative electoral votes, and new polls:
Colorado (Obama +7.8) 273 EV
Mason-Dixon 10/28-29 (9/29-10/1):
Obama 49 (44)
McCain 44 (44)
Virginia (Obama +6.6) 286 EV
Nevada (Obama +6.2) 291 EV
Mason-Dixon 10/28-29 (10/8-9):
Obama 47 (47)
McCain 43 (45)
Ohio (Obama +6.0) 311 EV
Mason-Dixon 10/28-29 (10/16-17):
McCain 47 (46)
Obama 45 (45)
Florida (Obama +3.3) 338 EV
North Carolina (Obama +2.8) 353 EV
Mason-Dixon 10/28-29 (10/23-25):
McCain 49 (47)
Obama 46 (47)
North Dakota (Obama +1.3) 356 EV
Missouri (Obama +1.2) 367 EV
Mason-Dixon 10/28-29 (10/22-23):
McCain 47 (46)
Obama 46 (45)
Indiana (Obama +0.6) 378 EV
Montana (McCain +1.8) 160 EV
Georgia (McCain +3.4) 157 EV
Arizona (McCain +3.7) 142 EV
West Virginia (McCain +7.4) 132 EV
McCain Base (127 EV): Alaska, Utah, Idaho, Wyoming, South Dakota, Nebraska, Kansas, Oklahoma, Texas, Louisiana, Arkansas, Mississippi, Alabama, Tennessee, Kentucky, South Carolina
The poll averages are created by a magic spreadsheet. Self-selected (Internet and mail) polls are ignored; no favoritism is done among the remaining pollsters. Polls are adjusted to today's conditions by shifting them by the amount of change in the average of Gallup and Rasmussen tracking polls. The weight of polls in the averages decreases geometrically each day such that 7 day old polls have 1/2 weight and 14 day old polls have 1/4 weight. The weight of state tracking polls is divided by the number of days in the sample. This method is very responsive to recent changes in both state and national polling.