Seth Flaxman, from the Oxford's Department of Statistics, explains how he has used data to enable predictive analysis of US elections.
By using ecological inference and census data, Seth has estimated not only “exit poll” style results (such as Trump’s level of support among white women), but entirely novel categories. The data analysis has also allowed researchers to explore which characteristics listed in the census predict voting behaviour, and non-voting.
The presentation is also available to download here.