The most recent CBS News/New York Times poll had Obama leading among likely voters nationally 52 percent to 39 percent, and the previous CBS News poll showed a similar lead. Guided by the results of the two polls, YouGov/Polimetrix then modeled 50 state electorates for CBS News, incorporating information from 26,671 online interviews with registered voters.
YouGov/Polimetrix applied the CBS News/New York Times double digit-lead to their 50-state national results weighted to each state's relative vote size, demographics, and matched to the CBS News/New York Times Poll in national party identification and ideology.
The technique allows the CBS News elections and survey unit and YouGov/Polimetrix to to estimate the probability that each candidate would carry any given state -- as of now -- in a national race. In late September, the national poll's 5-point lead for the Democrats translated in the CBS News/YouGov model into a most likely outcome of 313 Electoral Votes for Obama and 225 Electoral Votes for McCain,. In the current modeling, the most likely outcome produced by a double-digit Obama lead could be an Electoral College landslide, Obama 390 Electoral Votes, McCain 148.
That kind double-digit national margin would bring tossup states like Florida and Ohio into the Obama column in the model. North Carolina and Georgia would also move. Other battlegrounds like Nevada, Virginia, Pennsylvania, New Mexico, Minnesota, Colorado, New Hampshire, Michigan and Indiana, were part of the coalition of states Obama would hold with a 5-point national vote lead.
Also included in the model is what could happen to the Electoral College results if white voters became more Republican – or more Democratic – than election polls are currently suggesting.
Seemingly small percentage shifts in white voter support dramatically impact Barack Obama's overall Electoral College fortunes in the model. If Obama were to win even a narrow victory among white voters (which no Democrat has done since 1964) he would amass an electoral landslide of 408 Electoral Votes in the model. If Obama loses whites narrowly, taking 48 percent (as Bill Clinton did in 1996), it would still probably produce a comfortable Electoral College win: 379 Electoral Votes. In such a scenario Indiana and Ohio would become tossup states, with Indiana likely turning Republican. Getting 45 percent of the white vote would still give Obama a comfortable Electoral College win.
If Obama gets 41 percent of whites nationally, as John Kerry did in 2004, this analysis suggests he would lose narrowly in the Electoral College to John McCain, getting 247 Electoral Votes to 291 Electoral Votes for McCain. With that shift in national white vote, McCain takes Ohio and also becomes more likely than Obama to win North Carolina, Colorado, and Florida, among other states.
If Obama underperforms Kerry, taking just 39 percent of the white vote, he would probably lose badly in the Electoral College. McCain would amass a comfortable win and Obama would get just 209 Electoral Votes. Obama would win few of the current battleground states; McCain would probably capture Pennsylvania and Wisconsin, too.
How The CBS News/YouGov Model Works:
The national horserace percentage is not distributed evenly across all states. States differ in size (how much they contribute to a candidate's national vote total) registration, turnout, demographics, partisanship and, of course, candidate preference.
To model these state-level differences, we incorporated 51 additional sets of interviews with voters conducted by You Gov/Polimetrix, matched to voter registration lists, partisan and demographic profiles in each of the 50 states and the District of Columbia. Results were then adjusted to reflect the latest national CBS News/New York Times poll horserace estimate. Using all these data, each state was assigned a probability of going to either Obama or McCain.
The electoral tallies are a model of the Electoral College's most likely outcome if a candidate were to win the national popular vote by the margins in current polls. Because some states shape up as closer contests than others, a shift in national vote choice will mean a change in the likely winner in the close states more readily than in less competitive states.
You Gov/Polimetrix selected respondents for this modeling from its online panel of 1.1 million Americans. They were matched to a sample of registered voters from each state to ensure representativeness.
As the campaign goes on, the underlying pattern of results could change state by state. CBS News will continue to look at the state by state implications of the national polling numbers.