By Rebekah Castor Syracuse, N.Y. (NCC NEWS) — As the countdown to the presidential election continues, physicists at The City College of New York believe new analytic tools produce a more accurate method of predicting election trends with Twitter. Hernan Maske, Alexandre Bovet, and Flaviano Morone analyzed 73 million tweets collected from June 1 to September 1, 2016 to infer the opinions of Twitter users regarding the presidential candidates.
“We can produce with really high accuracy the NY Times polling,” Morone said. “We can also forecast by 6 to 15 days the results of the NY Times poll.”
This is the first time a social network can be used to validate election polls, according to Morone. The physicists got the same results as the NY Times, but for a fraction of the price. Because current polling methods are very expensive, Twitter and other social networks are the future for predicting outcomes with large data. Twitter could replace the multi-billion polling industry in the next election practically for free, Morone said.
“We agree with the big industry of polling and predict that Hillary will win when we take into account the whole population of Twitter,” Morone said. “This is the same prediction as the NY Times.”
However, when only analyzing influential Twitter users, Morone and his team predicted Trump will defeat Clinton. Influential users are those who Tweet consistently and interact frequently with other users, according to Morone.
“This is contradictory,” Morone said. “But the whole population of Twitter is a more accurate result than just analyzing influential users.”
When analyzing the whole population of Twitter, each user counts as one vote. Therefore, influential users may support Trump strongly, but on election day they only cast one vote. This analysis is more representative of the national population, according to Morone.