Dan GrossmanCan someone give some explanation why Intrade indicates Obama 59% likely to win and McCain 41%, while the polls show them about even?

Are there any other web sites you find useful in predicting the election?

Alex Forshaw replies:

The polls don't show them as even. Additionally, the most credible polls (Gallup, Rasmussen, ABC/WP) are generally the ones which show Obama with the largest leads.

Real Clear Politics has a summary of some of the major polls.

Tom Marks and Jason Ruspini add:

There are two methods to predict the election.

The prediction markets such as Intrade and Iowa Election Markets (chart) are probably the best predictor.  As of 10pm EST today Intrade is giving 57% chance for Obama and 43% for McCain. Iowa is also 53 vs 47.

Another way is to look at polls. However, the polls generally cited by news organizations are national polls which on the surface is fundamentally flawed since it is the Electoral College Vote and not the Popular Vote that determines the election result.

What you must do is look at state polls and infer the Electoral Vote implied by these polls.  Fortunately there are some very nice web sites that do the work for you.  They automatically look up the results of the latest polls and apply statistical adjustments (such as weighting polls differently depending on accuracy and timeliness, or performing Monte Carlo simulations to incorporate the inherent inaccuracy of any poll) to come up with their forecast.

The three best Electoral Vote prediction sites based on state polls are:

Vanderbilt U. Economist A. Moro's Forecast

Sam Wang's Princeton Election Consortium (chart )

Five Thirty Eight

Reading about the methodologies these sites use can be interesting for those who are statistically oriented.  Most people probably only care about the result.  Currently Prof. Moro gives 314 votes for Obama. Prof. Wang gives him 311 while FiveThirtyEight gives 309 to Obama. (As every schoolboy knows it takes 270 electoral votes to win the election).

In summary, both methods of prediction currently favor Obama. The rough equality shown by some national polls is misleading.





Speak your mind

5 Comments so far

  1. Jeff Watson on September 7, 2008 12:37 am

    I’m curious about the accuracies of the polls in this current election cycle. Ever since the right has been very energized by the nomination of Sarah Palin, there might be a surprise in store. Since I’m not a fan of surprises, I like to look outside the box to determine the likelihood of different outcomes. Which brings me to the Battleground Poll, and question D3.

    The Battleground Poll (a bipartisan poll conducted by the Terrance Group and Lake Research Partners)asks the same question in every poll. Question D3 asks,

    “”When thinking about politics and government, do you consider yourself to be…”

    Very Conservative 20%

    Somewhat Conservative 40%

    Moderate 2%

    Somewhat liberal 27%

    Very liberal 9%

    Unsure/Refused 3%

    This question has remained the same, in every one of their polls, and the percentage has changed little over the past decade. This question only shows up in the internals of the poll and isn’t represented in a chart of their study. I like the Battleground Poll because it’s a cooperative effort between 2 polling firms, one Republican and one Democrat. Both are in agreement with polling, questions, and sample sizes and methods.

    As a side note, I was intrigued by the small percentage of people that considered themselves to be moderate. However, I did note that the last answer, “Unsure/Refused” would give anyone an out that did not want to label themselves in the other categories. My only problem with this question is that they didn’t include Objectivist as a possible category.

    60% of people leaning towards Conservatism(according to this poll) could affect the result in the privacy of the voting booth.

    Although the following link doesn’t include the internals, it provides interesting study and a good read.

    Here is a link to the actual questions and internals in the poll.

  2. steve leslie on September 7, 2008 6:52 am

    I don't know much about polling but I know alot about people and one thing that a poll will never be able to predict is the quirkiness of the American Voter.

    Let me expand. If the weather is particularly bad that day many will not vote period. Our country is notoriously terrible at attendence. Only 50 percent at best show up for a national election. Therefore if a battleground state has terrible weather, that can have a big impact on attendence or participation if you will.

    An ephemeral moment. What shenanigans might go on during the day. In 2000 Dan Rather predicted Gore as the winner due to exit polls in Florida. However, the exit polls did not include the write-in voter which largely favored Bush and the military votes that favored Bush. Some have specultated that this was a veiled attempt to influence the election and to discourage those who vote in the panhandle where polls closed one hour later due to being in a different time zone from continuing to vote.

    Extending voting hours. I seem to remember Missouri and its Democratic Governor pulled a fast one here in 2000 citing a laundry list of reasons such as machine malfunctions, etc. and extending voting hours to help their cause.

    Long lines at the voting sites. We are an impatient group and if lines are long, many will not vote because they just won't stand in line.

    Which candidate is most effective in their 2 minute offense. Elections are decided by the middlers. These are the people who will not or cannot make up their minds until the last moment. Truman in 1948 Gore in 2000 and Kerry in 2004 are examples of this. So the candidate who is best organized and can put in the final big surge and leg kick during the home stretch oftentimes will make or break the final tally.

    What major news event may transpire in the final days or hours prior to the first Tuesday in November. Carter's inability to close the deal and resolve the American hostage situation in Iran showed him to be too dovish and Reagan used this to great advantage to win in 1980.

    A recount vote and many other thoughts and conjectures might be postulated but the point has been sufficiently made.

  3. Rocky on September 7, 2008 8:37 pm

    Let's not forget Ray Fair's Model for Presidential Elections. It's the Occam's Razor of predictions and his original model has been around since 1976. As of 7/31/08, he had the Republicans at 48.5%. model requires only three inputs….You can play with the variables at:

  4. Bob Johnson on September 8, 2008 9:01 am

    And this from today 9-8-2008; must be the Palin Bounce: "In his best Gallup poll showing in months, GOP presidential nominee John McCain enjoys a post-convention bounce and moves ahead of Democratic rival Barack Obama, 48 percent to 45 percent."

  5. michael bonderer on September 10, 2008 6:59 pm

    As of this afternoon, Intrade is now showing McCain odds just over 50% and surpassing Obama for the first time


Resources & Links