Nate Silver's work speaks for itself. His talent with election modeling has led to a new home at the New York Times. Therefore, we at Recorded Future could not help but notice his statement about momentum and this electoral cycle on Twitter this month. We have previously discussed in this space how to use Recorded Future's news analytics API to measure and track the positive and negative sentiment of the vocabulary surrounding political events. Recorded Future also is able to measure momentum, which we will look at here.
On our Senate Campaign Analysis Dashboard, we are continuously measuring the sentiment and momentum of the candidates in all the close 2010 Senate races. Momentum, in our calculations, is an attempt to measure online buzz -- it values not only the total frequency of media mentions but also the visibility of each hit and the credibility of the source, among other factors. Note that this is somewhat different than the way that Nate Silver discusses momentum. Still, we though it would be interesting to compare findings based on our momentum measure with findings based on his. Does an increase in online buzz lead to a more favorable outlook for the candidate's chances?
While there are many polls-of-polls, for this exercise in "shorting" a candidate, we will look to InTrade.com. InTrade is a global valuations market where users can make contracts on possible outcomes; a event's price point of $18 means a user would lay $18 to win $100 if the event were to occur and get back nothing if it did not. The higher the price, the more likely the market believes the outcome to happen. Right now, with an asking price of $60.90, a bettor would risk $60.90 to win $100 if a Democrat wins the 2012 presidential election and lose it all if a Republican were to prevail. And like all betting markets, the price fluctuates continuously as more information comes in and more people make contracts. We can examine the day-by-day valuations of candidates in the current Senate races to see if the global betting market reacts to momentum fluctuations as measured by Recorded Future.
The first Senate race examined, based on Mr. Silver's suggestion, was the one in Connecticut. I pulled from Recorded Future each day's aggregate momentum score over the past month for both Richard Blumenthal, the Democratic candidate, and Linda McMahon, the Republican. I also downloaded from InTrade the closing asking price on a bid to win $100 for both. These are the results:
The line graphs (blue for the Democrat Blumenthal and red for the Republican McMahon) show the valuation of the InTrade closing price. The columns show the Recorded Future momentum score, valued on the right vertical axis.
We can then count how often the InTrade value reacted to a candidate having online buzz. McMahon had greater momentum on fifteen days, and on the day after, six times the InTrade values moved in her favor, seven times it moved against her, and twice it stayed the same. Blumenthal had sixteen days with a higher momentum score, and following those, InTrade moved with him eight times, against him six times, and twice did not change.
We can also consider the large magnitude changes in momentum. Five days McMahon's momentum was more than twice Blumenthal's. Of those five, twice the market moved in her favor, twice against her, and it did not change once. Of Blumenthal's four days in which he more than doubled McMahon's momentum, we see that the market moved in his direction once and against him three times.
Our findings then correlate with Silver's: there does not seem to be a strong signal between an increase in a candidate's momentum and the market willing to bet on that candidate to win. To test this, we can compare the results to other races by running the same momentum/InTrade analysis. Here's California, where the Democrat Barbara Boxer is facing the Republican Carly Fiorina: