Will April's Facebook 'likes' predict the outcome of May's general election?

Using Facebook to express support for politicians and their parties is standard practice for voters these days – but does how many 'likes' a party has before an election have any bearing on the eventual results? A new study of the 2014 Indian election published in the Asian Journal of Political Science suggests that it does. Knowing that "clicking the 'like' button on a Facebook fan page could be considered as an expression of liking or an act of lending support", Francis P. Barclay from the PSG College of Arts and Sciences in Tamil Nadu and his colleagues analysed the number of 'likes' the leaders of the Congress, BJP and AAP parties received on their fan pages in the days leading up to the May 2014 Lok Sabha elections.

Based on previous studies, Barclay and his colleagues expected there to be a correlation between 'mass political preference' and 'likes' on fan pages. "It is also logical to assume that [the] more people there are that prefer a party, the more votes it will secure in the election. So, it is safe to assume that mass political preference has a direct positive effect on the number of votes."

They hypothesised further: "If there is a strong correlation between the number of 'likes' recorded during the election campaign period and the vote count, then the former could be used to predict the latter. In other words, the politician or party that secures the most number of 'likes' on Facebook during the campaign period is likely to secure the most number of votes in the elections."

Once all the data were collected and several variables taken into account, Dr Barclay and his team crunched the numbers. The result was a powerful regression equation able to predict the probable winner.

They concluded: "When the percentages of 'likes' recorded on the official Facebook fan pages of the political parties are correlated with the election results, a linear and positive association is found."

Indeed, when the researchers used data from the month before the election, "the probable vote share of these parties can be deduced and the winner can be predicted with an accuracy of 87%"; when data from the several months before the vote were taken into account, that accuracy dropped slightly, to 69%.

Although Barclay and his colleagues only studied one Indian election, their results have important implications for UK politicians, journalists and opinion pollsters as our own general election approaches. Their work has established beyond question that Facebook is an accurate guide to the public's current mood – and where that mood will take them in the polling booth.