Back to questionnaire

Electoral vote

Grey states have less than 10 votes. We need at least 10 voters to show a result for a given state. Share the survey among your friends to find out who will win your state!

Electoral votes over time

The graph shows the total electoral votes for each candidate across time, based on the estimated responses from our survey participants. It is updated on a daily basis until November 8th.

The light blue and light red areas describes the average error. This will decrease as we get closer to election day and as the sample size increases, yielding a more and more precise estimate of the final vote share for both options.

Total votes

Total of users voted. These do not represent our prediction, just the total distribution of current votes.

Are you happy with your prediction? Then share it with your friends, and attain bragging rights on election day. Remember, if it's not on Twitter, it didn't happen!

Would you like to share to your friends who you voted for?

Results for your Twitter followers

Total of users came to this application trough your tweets or shares. That ranks you by influence on other users

Total of of your Twitter followers voted.

How did your followers vote? Do you all vote the same way?
Share if you are happy with how they voted?

Results for the Twitter users you are following

Total of people you are following voted in this application.

How did the people you follow vote? Do you all vote the same way?
Share if you are happy with how they voted?

The obvious difference in the standard polling approach and our BASON method can be seen in the graphs above. The standard polling method produces a very biased estimate in a small sample that yields the current distribution of votes to be, for example, 65% for Hillary and 35% for Trump overall. This is obviously wrong and unrealistic.

However our prediction approach does not suffer from a biased sample nor does it suffer from the self-selection problem. Our prediction method utilizes the wisdom of crowds approach in a novel fashion.

By asking our respondents not only to express their preferences, but also who they think will win and how they feel about who other people think will win, we can use even a modest sample to make precise estimates of the likely outcome. This is the benefit of group level forecasting. Our diverse, decentralized, and independent group of respondents has given us very realistic estimates of the outcome, meaning that the final estimate is likely to be quite close to the actual outcome on November 8th.

Users of the app over time