Experts explain why 'Big Data' got it Wrong in Chilean Vote

Those with an interest both in Latin American affairs and also the use of data to make political predictions might have noted that forecasts of the result of the Chilean Referendum late Sunday. Results based on Big Data didn’t turn out as expected.

Chile is one of the most economically and socially stable countries in South America. Without dwelling on the detail of the vote which was whether to accept a new, more liberal constitution, the pundits had correctly predicted a win for ‘reject’. The final result was however far higher than had been expected. Some analysts who got it wrong were even further off the mark by suggesting that the new constitution would be approved.

It was particularly interesting to hear their explanation, which was principally based on the impact of social media analysis. They said that because the vote was mandatory rather than voluntary, then the prediction model was skewed as the vote involved all sectors of society rather than just those who were 'connected' on social media. As a result, predictions based on social media analytics alone were unrepresentative of the population as a whole, and therefore had an inherent bias.

Professor Oscar Jaramillo of the Finis University School of Journalism said;

'There is a misconception, and that is to consider that social networks are public opinion….?Basically what social networks can tell us is which sectors are more active than others, but they do not represent in any way what we could call the silent majority.’

Maria Jaraquemada, Director of Transparent Chile, also commented on the error, saying;

‘Clearly, what people were expressing through social networks and in the Google search is not something that, at least in this case, could be extrapolated to the rest of the population…. The fact that we search for keywords in Google, for example, has to do with the controversies which has occurred around these issues in previous days, more than really a matter of electoral preference.’