Follow us on our previously unused Twitter account @claritycampaign for publicly available projection sites in the future.
As a research and development exercise for future real-time modeling applications, we built an online dashboard that predicts final statewide results based on partial actual results . Basically it takes the actual results as they are reported, and uses a constantly self-adjusting model to project the final result statewide based on the results reported thus far (taking into account geographic differences, turnout and performance). This is essentially what many media outlets have staff doing behind the scenes in order to ‘call’ races but here you can follow the results in real time and make your own judgement based on the model.
NOTE, that as a starting point we mocked up each race we’re tracking (regardless of current polling) as a 50/50 split between candidates – so in races that are not going to be close we’ll quickly see the projections divert from the starting scenario once real data is introduced.
To put it simply, this dashboard will put the actual results into context – a campaign might be in a situation where they are up 10,000 votes with 30% of the precincts or counties yet to report. This application will tell us if that 10,000 vote is good or bad, based on the predicted partisanship of those areas that are yet to report.
We deployed custom precinct-based dashboards for several clients internally but public ‘beta’ versions of county-based results and projections may available in certain races and states as a proof of concept (predictions will be less reliable than precinct-based sites).
Here is a snapshot from the Kaine for Senate boiler room on Election Day 2012 in Virginia (congratulations to an incredible campaign team!).