Forecasting 2014 Congressional Elections Using Facebook Metrics

American vote buttons illustration

by Matthew C. MacWilliams

Facebook constantly tracks the growth of each candidate’s fan base as well as the number of people engaging directly with candidates online. These Facebook metrics comprise a rich dataset. Can they be used to predict the outcome of campaigns for U.S. Senate? An exploratory investigation in this issue of PS: Political Science & Politics by Matthew C. MacWilliams from the University of Massachusetts, Amherst finds that Facebook fan likes and PTAT statistics, when added to electoral fundamentals similar to those used by political scientists in national-election forecasting, produced surprisingly accurate predictions of winners and losers in the seven hotly contested 2012 Senate races studied. Additionally, in five of the eight weeks leading up to the 2012 general election, the simple Facebook Model more accurately predicted final election results than poll-of-polls averages. The question remains, however, whether the results are an anomalous “flash in the 2012 pan” or an indication that using Facebook metrics to measure campaign effectiveness is a new tool scholars can use to predict the outcomes of congressional campaigns.

Forecasting 2014 Congressional Elections Using Facebook Metrics, PS: Political Science and Politics, by Matthew C. MacWilliams / Volume 48 / Issue 04 / October 2015, pp 579-583