Analytics abuzz

Much has happened since the last issue of OR/MS Today, including the re-election of President Obama and the presentation of the 2012 INFORMS Annual Meeting, and both events had analytics fingerprints all over them.

The conference, which drew 4,500 or so attendees to Phoenix, was buzzing with analytics, from the special sessions on the INFORMS’ new Certified Analytics Professional (CAP) program to the fly on the wall in the room where the INFORMS Board planned the next steps they’ll take to further establish INFORMS as the world’s “premier organization for advanced analytics professionals.”
INFORMS President-Elect Anne Robinson has spearheaded the analytics movement within INFORMS from the time she first joined the board five years ago. During the conference, I caught up with her for an-hour long interview, and not surprisingly, we spent a good deal of the time discussing INFORMS’ embrace of analytics and the opportunities it presents for the Institute and its members (see “Ambassador of Analytics” on page 40).

While INFORMS carves out its place in the analytics space, it remains renown in the O.R. community for its world-class journals and conferences, and perhaps no organization does a better job of handing out awards. To see who won what this year, check out pages 48-53, followed by a conference photo essay on pages 54-55.

Among the thousand or so technical sessions presented at the conference, one that caught my eye was titled, “United States Presidential Forecasting: Who Will Win the White in 2012.” Chaired by Sheldon Jacobson, the session included a presentation by quantitative historian Allan Lichtman on the “Keys to the White House.” Lichtman’s 13 keys have “retrospectively accounted for the popular-vote winners of every American presidential election from 1860 to 1980 and prospectively forecast well ahead of time the popular-vote winners of every presidential election from 1984 through 2012.” Using his “keys,” Lichtman predicted an Obama victory two years before the 2012 election. The session in Phoenix came shortly after Obama’s disastrous performance in the first debate and three weeks before the election when national polls showed a virtual tie and momentum switching to Romney, but Lichtman stood by his prediction.

For his part, Jacobson outlined “a Bayesian approach incorporating undecided voter effects into the analysis” to predict the number of Electoral College votes each candidate would receive. The results, taking into account a combination of polls in each state, were updated daily and posted on his “electionanalytics” Web site. Throughout the summer and fall, the numbers showed Obama with a huge probability of winning the most electorate votes and thus the election, culminating with a .999 probability on the morning of the election.

While analytics clearly helped the mathematical analysts (as opposed to the political types) make better predictions, it also helped the Obama campaign mount an effective ground game that identified Democratic voters in swing states and used analytics to deliver micro-targeted messages and devise an effective door-to-door effort to boost turnout on Election Day. Doug Samuelson reports (page 64) the Romney campaign also developed an analytics-driven, get-out-the-vote effort, but something went awry on Election Day. Blame the quants, not analytics.

— Peter Horner, editor