Today we welcome back UMassAmherst political science doctoral candidates Matthew MacWilliams, Edward Erikson and Nicole Luna Berns and their Facebook Forecast Model for an updated prediction on the Brown-Shaheen race and an important challenge Brown poses for their model.

This week’s forecast shows Senator Jeanne Shaheen still ahead of Scott Brown in New Hampshire, but Shaheen has dropped from 55% to 52%. The Scott Brown candidacy in New Hampshire presents an interesting outlier challenge to our Facebook Forecasting Model. In fact, there is no other candidate in the history of Senate campaigns quite like Brown. Why?

While Brown is not the first candidate to run for U.S. Senate in more than one state, in fact James Shields was elected to the Senate from a trifecta of states – Illinois, Minnesota and Missouri, Brown’s candidacy is still an outlier in the history of the U.S. Senate. What makes Brown’s peripatetic political travels even more of challenge to our forecasting model, and his candidacy a campaign singularity, is that his Facebook “likes” do not recognize state boundaries and travelled with him to New Hampshire. Thus, the day in April he announced his challenge to Senator Jeanne Shaheen, Brown’s Facebook “likes” stood at 372,011 and his “PTAT” logged in at 3,417.

Certainly, on the day he announced, most of Brown’s “likes” and “PTATs” were not New Hampshire voters who had jumped on his bandwagon. These “likes” and “PTATs” were the vestigial remnants of his two campaigns for Senate in Massachusetts. And while our model is specifically designed to track the growth of “likes” and fluctuations of “PTAT” for candidates running in one state and takes a ceteris paribus approach to out-of-state Facebook fans and engagement with candidates, Brown’s multiple state base of “likes” and “PTATs” is theoretically and practically different and may bias our forecasting model against his candidacy.

To account for the outlier nature of Brown’s candidacy while maintaining the integrity of the theoretical approach of our model, we have also tracked the New Hampshire Senate race using “like” and “PTAT” adjusted for Brown’s base on the day he announced. This week we report our forecast for the New Hampshire Senate based on both original and adjusted data at www.hashtagdemocracy.com.

The forecasting models based on both original and adjusted data show Senator Shaheen still in the lead. In the original model, the race has tightened some with the three-week, rolling forecast predicting a Shaheen taking 55% of the vote. The adjusted model shows Shaheen with a much smaller lead over Brown. The two-week rolling forecast predicts a Shaheen victory by just 52%.

For now until Election Day, we will report both original and adjusted forecasts. Only time will tell whether or not our model can accommodate the Curious Case of Scott Brown.