IPPSR Outlook ’08 has been tracking the polls and pollsters as they measured the likely popular vote nationally since this past summer. Outlook '08 has identified battleground states and wrapped poll results into a mathematical projection of Electoral College potential. Outlook '08 is not meant to predict an outcome, but to offer an analysis focusing on the Electoral College and its importance in the election of a U.S. President. It is an outgrowth of work to update a 2004 IPPSR Policy Brief, Electing a U.S. President (pdf.).

Presidential Election Results
Tuesday, Nov. 4, 2008
Polling as of Monday, Nov. 3, 2009

Popular Vote Results
U.S. Sen. John McCain, Republican, 46%
President-elect Barack Obama, Democrat, 52.7%

Electoral College Results
U.S. Sen. John McCain, Republican, 162
President-elect Barack Obama, Democrat, 365

Total Electoral College votes needed to win: 270
Total Electoral College votes: 538


Battleground States

IPPSR's Outlook '08 analysis is unique for its transparent definition of battleground, or toss-up states. It is also unique in that it breaks from tradition in identifying battleground states by historic voting patterns, and relies instead, on current voter sentiment. IPPSR's Office for Survey Research defined such states on the basis of narrowly divided electorates and a high percentage of undecided voters. More than 400 individual statewide and national polls were reviewed as part of IPPSR's Outlook '08 methodology.

More precisely, IPPSR battlegrounds are those states in which the percentage of respondents who say they are not voting for either of the major candidates is more than twice the difference between those who say they intend to vote for one of the two major candidates.

Post-Election Analysis

After five months of research, starting on June 7th, the day that the race for the presidency officially became a two-person contest, we have a conclusion to the contest and we have reached some conclusions about the process of political polling.

With this election, polls proliferated in the rush to gain an understanding of the electorate and perhaps an edge on the opposition. There were two fundamental differences in how IPPSR organized the daily snapshots of the race into Outlook ’08 and how other researchers treated similar information.

• First, Outlook ‘08 did not report averages or trends in assessing the likelihood that any given state would be declared for one candidate or the other. IPPSR’s research method in Outlook ’08 was to treat the most recently reported poll as an election return. For example, when more than one state was polled on a given day, Outlook ‘08 averaged the percentages for each candidate and reported the result. A tie was declared a tie.

Other researchers may treat the information differently. They may have averaged the most recent results or used historical data, for example from previous presidential races, to arrive at the most likely Electoral College tally.

• Outlook ’08 defined a battleground state as any state where the percentage of respondents declining to vote for either Sen. Obama or Sen. McCain for any reason -- indecision, intention to stay at home on election day or to vote for a third party candidate –  was at least twice as great as the margin of difference between Obama and McCain. This approach called upon real-time information, that most recently available, rather than relying on past election histories. As a result:
  
• States entered and left the Outlook ‘08 battleground
  
• A 45 percent to 40 percent difference (5 percent) between the two candidates would denote a battleground as it signified mathematically that 15 percent of the respondents were not voting for either candidate. The computation: 15/5>2.0
  
• A 51 percent to 48 percent (3 percent) margin was deemed likely, because only 3 percent of the respondents were identified as undecided

Outlook ‘08’s final predictions, as of noon on Nov. 5, 2008, missed only two contests. The final Outlook '08 poll, reported on E-1, Nov. 3, 2008, showed Obama winning in Montana and McCain taking Indiana. Aside from that, the final polls predicted the actual outcome.

As of the time of this publication, North Carolina and Missouri, battleground states in the pre-election update, had not been decided for either candidate.This is evidence of the accuracy of the pre-election polls. Outlook 08’s Electoral College count for Obama settled at 367. In actual result, Obama's Electoral College count rested at 349, with 26 votesleft undecided at the time of this publication.

During the final three weeks of polling, the estimates of popular vote averaged from 50 percent for the Obama-Biden ticket to 44 percent for the McCain-Palin ticket. In the final counts, the election went 52 percent for Democrats, 46 percent for Republicans, indicating a dead-even split of the undecided, and an identical 6 percent margin of victory.

As a final comment, discounting of political polling has been a popular pastime. Modern political polling offers sophisticated methodology. It also employs protocols meant to assure accuracy. These protocols can differ vastly among the pollsters who rely upon them. Some pollsters use automated voices to qualify voters (as "likely," for example) and thus to collect opinions. Some automatically discount voters who have not voted previously as "not likely" to vote at all.

Just in the last week, one pollster reported a one-day -- in the midst of a three-day rolling poll – 2 percent advantage for McCain among voters deemed likely to vote when the total response showed a 10 percent advantage for Obama.

Direct source material, including the entire questionnaire with all of its wording, the respondent qualifications, and adjacent questions will form the most reliable basis from which to draw conclusions about differences in political polls.

Before we exit, a word about exit polls and their major shortcomings:

• When asked, self-identified Democrats or Independent respondents are 50 percent more willing to talk to an exit pollster than are Republicans
• A substantial proportion of voters – at least 30 percent – now cast ballots in advance of an election. Early voting may take the form of absentee ballot or appearance at in-person polling places prior to the election day. The state of Oregon has mail-only voting.
• The methodology of sampling and weighting exit polls is highly suspect. These methods must be reviewed before reliable conclusions can be drawn.

2008 Campaign Perspective
Obama has been in the lead in the 2008 presidential election since IPPSR began tracking the polls this summer, except for a time immediately following the Republican Party convention. McCain outpaced Obama for six days --immediately following the GOP convention -- by an average of 1.5 percent.

To put the 2008 race in perspective, George W. Bush had a 2-point edge in the popular vote in 2004, and a 1-point deficit in 2000. Correcting for a two-person race, Bill Clinton defeated Bob Dole by 9 percent in 1996 and George H. W. Bush by 7 percent in 1992. The elder Bush defeated Dukakis in 1988 by 8 percent, and Ronald Reagan won over Walter Mondale by 18 percent in 1984.

2004 Presidential Polling History
The final polls of 2004 put President George Bush ahead of Democratic challenger John Kerry by 222 to 207 Electoral College Votes.

Another 109 Electoral College Votes came from nine states considered "battleground," defined as toss-ups as polls there found less than 2 percent difference between the two candidates.

Those states were: New Hampshire with four votes, Pennsylvania and its 21 votes, Ohio where 20 Electoral College votes were up for grabs, Wisconsin with 10, Minnesota counting 10, Iowa with seven votes, Nevada with five, New Mexico with five and Florida where 27 Electoral College votes could go to either candidate.

In actual voting, the two candidates split the nine states 5-4. New Hampshire, Pennsylvania, Wisconsin and Minnesota went to Kerry, who won 252 total Electoral College votes. Bush won Ohio, Iowa, Nevada, New Mexico and Florida and the White House with 286 Electoral College votes.

Notably, the pre-election polls in the states considered solid for each candidate accurately predicted that state's actual winner.

A Note On Polling Margin of Error
“Margin of error” is a statistical concept, and technically means that sampling error would produce a statistically significant result 5 percent of the time.

For example, if the actual margin between Obama and McCain is 6 percent, and the margin of error is computed at 3 percent, then 19 times out of 20, the difference between the two candidates would be between somewhere between 3 percent (the result of 6 percent minus 3 percent) and 9 percent (6 percent plus 3 percent). 

This computation applies to those polling results from THAT DAY, at that very point in time. The results are a like a photographic snapshot, and essentially freeze public opinion at a single moment. 

A point to remember: The latest polls are the best measure of what is the true score right now. Prediction of the future remains a guess, not a guarantee.   

--Nat Ehrlich, Senior Researcher, Office for Survey Research, Institute for Public Policy and Social Research in the College of Social Science at Michigan State University.

Thank you for your interest and for your attention in Outlook '08.

Your observations and input are welcome and may be emailed to Senior Researcher Nat Ehrlich at ehrlichn@msu.edu or to IPPSR Educational Specialist Cynthia Kyle at kylec@msu.edu.

Press Release (Aug. 4, 2008) (.pdf)
Press Release (Aug. 12, 2008) .(pdf)
Press Release (Sept. 17, 2008) (.pdf)
Changes in Electoral Vote, 1981-2010