In this year's Dutch general election, polls based on social media analysis, especially BuzzPeil 's , had a serious position next to the traditional opinion polls of Maurice de Hond and Ipsos Synovate . We saw on the evening before the election that both forms of opinion polls predicted roughly the same distribution of seats.
poll-guide-2012-09-11
Compared to the outcome, the polls were quite good, especially when taking into canada telegram data account the uncertainty margins of about 10%. The overwhelming victory of the VVD was not entirely predicted (varying from 34 to 37 seats, this became 41) and the SP was overestimated (varying from 20 to 22 seats, it became 15).
All this raises a number of questions. How do polls based on social media analytics work? How can they be so ‘good’ at predicting and why is there so much agreement with traditional opinion polls, which are created in a completely different way?