Post-Debate, the President-Elect Would Be...


As President Barack Obama and Governor Mitt Romney square off against each other tonight in the first of their three debates, the airwaves will be abuzz with pundits analyzing what each says and how that might affect the election outcome. You can judge for yourself or follow along with the analysis, but I, for one, will wait for the hard numbers to tell me who bested whom.

For surely the polls will tell that tale.

To get my data, I'll be heading to Election Analytics, self-described as a "Web tool that tracks and analyzes polling data to forecast who will win the upcoming November 2012 elections." As I write this post, looking at the latest available data, for yesterday, Oct. 2, I can see that if the election had taken place a day ago the Obama team would probably still be raising champagne toasts at the moment. Any way you slice the data, and even if swing states vote heavily Republican, the tool forecasts Obama as the victor.

The infographic below shows one slice of the data, for five days. You can also see results in histograms, sliders, and over longer periods. Election Analytics is tracking Senate races, too.

The data is only as good as the last poll results -- so this could change come tomorrow, post debate, as Sheldon Jacobson, co-creator of Election Analytics, told me in a phone interview last week. He recounted the 2008 election, when in mid-September the Election Analytics tool showed candidate Sen. John McCain having a sizable lead over Obama. "But by the end of September, that lead was gone and McCain never caught back up," according to the tool results.

That's to say, as election day draws closer, the Election Analytics forecasts turn into actual predictions. "Things start to get really interesting three weeks before the election -- that's when the rubber meets the road," said Jacobson, who is a computer science professor with a specialty in operations research at the University of Illinois in Champaign-Urbana.

The tool becomes most valuable heading into the election, he noted. As the polls change, Jacobson and his team "synthesize and analyze and run the data through our models." Those models, as demonstrated in 2008, seem pretty solid: Election Analytics accurately predicted results in every state but Indiana And even that didn't surprise the team, he told me.

"We looked at the tough states to forecast -- we had a group of really, really close states -- and we knew we'd get it wrong on one of them. We knew one of the states would flip, but we didn't know if it'd be, say, North Carolina, or Missouri, and it ended up being Indiana."

So how, you're probably wondering, as was I, does Election Analytics get it so right? The key is dynamic programming and sophisticated algorithmic techniques, said Jacobson, without revealing too much, of course. Here's a bit of info from the site:

The mathematical model employs Bayesian estimators that use available state poll results (at present, this is being taken from Rasmussen, Survey USA, and Quinnipiac, among others) to determine the probability that each presidential candidate will win each of the states (or the probability that each political party will win the Senate race in each state). These state-by-state probabilities are then used in a dynamic programming algorithm to determine a probability distribution for the number of Electoral College votes that each candidate will win in the 2012 presidential election. In the case of the Senate races, the individual state probabilities are used to determine the number of seats that each party will control.

You can read more about the methodology, including how polling data for each state is weighted and how Election Analytics takes into account swing scenarios, here. In addition, if you happen to be attending the Institute for Operations Research and the Management Sciences (INFORMS) annual meeting in mid-October, you can catch Jacobson in person as he'll be presenting on US presidential forecasting.

And, of course, be sure to head back to Election Analytics tomorrow, and then again following the Oct. 16 and Oct. 22 debates, as they're sure to shift the polls, Jacobson said. But back to today, the tool shows Obama would have won an election held yesterday with a 1.000 probability, capturing 342.2 electoral votes.

Do you trust Election Analytics results and other forecasts? Share below, and take our quick poll on the subject at the right.

Beth Schultz, Editor in Chief

Beth Schultz has more than two decades of experience as an IT writer and editor.  Most recently, she brought her expertise to bear writing thought-provoking editorial and marketing materials on a variety of technology topics for leading IT publications and industry players.  Previously, she oversaw multimedia content development, writing and editing for special feature packages at Network World. In particular, she focused on advanced IT technology and its impact on business users and in so doing became a thought leader on the revolutionary changes remaking the corporate datacenter and enterprise IT architecture. Beth has a keen ability to identify business and technology trends, developing expertise through in-depth analysis and early adopter case studies. Over the years, she has earned more than a dozen national and regional editorial excellence awards for special issues from American Business Media, American Society of Business Press Editors, Folio.net, and others.

Midmarket Companies: Bring on the Big Data

The "big" in big data is no reflection of the size of the organization embracing its potential.

Push Yourself to New Analytical Discoveries

Take inspiration from Christopher Columbus as you pursue your analytical journeys.


chicken or egg
  • 10/3/2012 4:30:50 PM
NO RATINGS

Which came first: A person's decision to vote a specific way or the results of the poll that influenced him?

Re: chicken or egg
  • 10/3/2012 4:52:37 PM
NO RATINGS

@Noreen, scary question. I would hope the former -- that is one reason we host public presidential debates -- but I suspect sadly plenty of people merely follow the crowd.

Re: chicken or egg
  • 10/3/2012 5:50:21 PM
NO RATINGS

Beth, - In a way i think polling houses shape the polls. Often though not always i've seen them use samples that are statistically not fair and draw conclusions about leading candidates. The results is that people's minds are influenced in that way and in the end it is a vicious circle. No wonder the politician who's not leading in opinion polls will always cry foul.

Re: chicken or egg
  • 10/3/2012 5:57:24 PM
NO RATINGS

This is one of the reasons the predictions are more accurate as you get closer to the event.  1) they become self reinforcing and 2) there's less time for mass changes in distributions to occur.  IF you were to plot the probability that their prediction is correct verses distance from the event I'll bet you'd get a nice smooth line (or at least a curve).

1.000 probability
  • 10/3/2012 6:24:18 PM
NO RATINGS

Beth,

Fully agree - the data is more interesting than all the pundits and "talking heads".  (Maybe we think that way because we're the type that likes an analytics site?)

One objection about the Election Analytics projection - it's too early to put a 1.000 probability on any of the elections.  Maybe 0.99, but not 1.000 just yet.

PC

Re: 1.000 probability
  • 10/3/2012 8:37:32 PM
NO RATINGS

PC, the 1.000 probability works, though, since the tool forecasts the winner based on that day's poll data. If it was predicting an outcome, I would agree, it's definitely too early.

Bayesian Estimators: Who Knew ?
  • 10/3/2012 8:57:26 PM
NO RATINGS

Thank you Beth for exposing a really unique election polling tool.  I don't believe much in polls, I mean if you are the one ahead then you can take some comfort, I mean it is better than being behind right ?

But when it comes down to it, I really don't think for the most part people know what they will do until it is actually time to decide. But I do find this application fascinating and this interesting mention: 

" The key is dynamic programming and sophisticated algorithmic techniques...."

Well I am certain it is  - fascinating use of Bayesian estimators as it applies here but I can respect their need for competitive privacy.

 

Re: 1.000 probability
  • 10/4/2012 4:24:43 AM
NO RATINGS

@Beth,

So it's okay to "forecast the winner"

but it wouldn't be okay to "predict an outcome"

Got it.... wait. What's the difference?

PC

Re: 1.000 probability
  • 10/4/2012 8:46:50 AM
NO RATINGS

@PredictableChaos -- no difference between forecasting the winner and predicting an outcome. The distinction I meant to call out is between today and Nov. 6. In other words, the Election Analytics tool tells you who would win if the election took place on the day for which it has the latest poll data. So looking at Election Analytics this morning, we know that if the election had been held yesterday, Obama would be the victor, earning 338.4 of the electoral votes, down slightly from the 342.2 electoral votes he would have earned had the election been held the day before, on Oct. 2. We'll have to see how the numbers shift once Election Analytics takes post-debate poll data into account. As the election date draws closer, and especially after that third and final debate, you can read the results more predictively -- in other words, if the tool shows Romney ahead on Oct. 28 that's a sign he'll be besting Obama come Nov. 6. Did I make sense this time? ;-)

Re: 1.000 probability
  • 10/4/2012 8:53:46 AM
NO RATINGS

Yep. Now I understand the difference. Thanks!

Page 1 / 4   >   >>
INFORMATION RESOURCES
ANALYTICS IN ACTION
CARTERTOONS
VIEW ALL +
QUICK POLL
VIEW ALL +