Your remark refers to the interesting example of Nate Silver. Predicting the result of 50 elections is a difficult problem indeed, but this technical difficulty was not reflected in the overly confident tone of political analysis.
Despite the 28.6% chances of Trump to become president, he still wrote a confident ‘Most Of outcomes Come Up Clinton’
Throughout the election, our forecast models have consistently come to two conclusions. First, that Hillary Clinton was…fivethirtyeight.com
I had a quick look at his methodology, which is pretty complicated:
We've just launched FiveThirtyEight's 2016 general election forecast, which projects how the 538 Electoral College…fivethirtyeight.com
Despite its complexity, his model does not seem to integrate the populist factor, which is new in America at such level of intensity.
As I understood it, the inputs of his predictive model are: polls, economic situation, and previous electoral outcomes.
But this does not seem to take into account the intensity of peoples’ political preferences, their level of passion, which was remarkably high for Trump supporters, and low for Hillary supporters. In turn, this passion impacted voter mobilization.
In order to impact the statistics on which Silver’s model is based, it would be necessary to have more frequent populist candidates to American elections. I think we should not wait until a populist president makes a disaster to revise the model.
The alternative solution that I sketched is:
- Get a better understanding of event attendance, how it is impacted by passions, using data from millions of Facebook events, with my app www.showup.tech
- Apply this model of event attendance to the prediction of voter mobilization, and integrate this result to Silver’s model.
I hope that the unfortunate result of this election will stimulate research on this topic.