Shortly after polls closed at 7PM on election night November 6, 2018, political campaigns across the country scrambled to predict the outcomes. Up to the time polls closed, campaign staff already had days to analyze a combination of early in-person votes and absentee vote-by-mail results as provided by the Georgia Secretary of State’s office, applying their best guess as to how each of those voters leaned. By 7PM election day, polls were closing and printed tapes, similar to adding machine tapes, with election day results were being posted on exterior windows and doors at each of Georgia’s 2,500+ polling sites.
One campaign that we were fortunate to work for, Sally Harrell’s campaign for Georgia Senate District 40, used an approach which enabled them to accurately predict the outcome of their election just hours after closing. This is an account of the process we used.
In order to predict the votes cast by each of the early in-person voters and absentee vote-by-mail voters, the campaign pulled together information from numerous sources, including National Party scores and personal knowledge gathered from interactions over the course of more than a year of voter outreach. To accomplish this sizable data collection effort across a voting population of more than 100,000 voters in the district, we deployed a disciplined data collection plan which leveraged technology and numerous tools to input and consolidate data from volunteers and other allies. The campaign staffed a dedicated volunteer outreach resource to lead volunteer organizing and communications for more than 100 dedicated regular volunteers and to support and train occasional volunteers. This effort produced voter intelligence data of fairly high confidence for most of the district’s voters.
Access to daily early vote counts from Georgia’s Secretary of State’s office is available to the public. From their Elections site, records indicate who has voted and by what method they voted (by mail or in person) but records do not indicate who or what anyone voted for. That prediction is reliant on the data collection effort described above.
By joining together the “who voted” data from the Secretary of State with the “how they likely voted” data amassed by the campaign, the team had a good idea of Sally’s position going in to election day.
Election night vote counts by candidate are available to the public as soon as the polling site completes their closing process. The challenge is collecting the results because of the state’s method of communicating results. In Georgia, each polling site posts a paper tape with vote counts (for all races on the ballot) on their front door. Some counties post a single tape for the entire polling site while others post a separate tape for each machine within the polling site. In order to calculate an election day winner, campaigns must send volunteers to each polling site in their district to sum up outcomes. Many campaigns don’t bother to do this. Those that do have a significant intelligence advantage early in the evening on election night. Sally Harrell’s campaign sent volunteers to all of the 47 precincts in the district. Volunteers took photos of vote counts for the State Senate race and reported them back to campaign headquarters.
Before election day, we believed we were solidly ahead based on analysis of early voting trends. Sally’s campaign manager requested that we develop a visualization to display at the election night party so that Sally’s supporters could watch the results roll in as we collected them. The dashboard we developed started off with our estimate of the early vote outcome. As election day results were submitted by campaign volunteers, we displayed them on the screen and showed the crowd how many more votes we believed remained unreported. The crowd of supporters was excited to be in the loop with the campaign and we were glad to see our expectations play out in real-time on the big screen.
At 9PM we had collected results from enough precincts to feel like the early vote momentum would not be reversed by election day voting. By 10PM, nearly all precincts were collected and the campaign had a solid idea of the election outcome. By 11:30PM, when the Secretary of State’s office had published enough results to confirm the collected data, we concluded the race was over and I was the lucky one who got to tell Sally that she won. Results were not reported in the press until the following day. It was an excellent experience that we will be repeating in future elections, including Sally’s run for reelection in 2020.