I read the recent Times article about election predictions going wrong
with great interest. For me the big loser in the election was not
Clinton (or Trump), but data science -- social science data science, in
particular. The reality is despite all the wonders that we've heard
about how one can do big-data mining on social science data
sets, finding, for instance, highly specific pockets of voters, the
polls were consistently wrong. Consistently, the Times predicted that
Clinton would easily win -- never, in fact, predicting the opposite.
The reality of election day was completely different, with Trump
easily trumping. I do not believe that there was a dramatic reversal
of opinion on the day of the election. The blunt fact is that much of the
earlier polling and associated reporting was inaccurate.
Unpublished letter in response to:
News Outlets Wonder Where the Predictions Went Wrong - The New York Times
http://www.nytimes.com/2016/11/10/business/media/news-outlets-wonder-where-the-predictions-went-wrong.html
Also see:
https://twitter.com/markgerstein/status/797462057134202880
