Although Ossoff did remarkably well among early voters in the April 18 runoff election, he performed much worse among Election Day voters, and many more voters went to the polls rather than vote in advance. This is also consistent with patterns of early voting in key swing states that looked advantageous for Clinton in the 2016 general election but did not provide a solid enough lead to win.
The final result in the runoff election underscores an important caveat when interpreting early votes: Election Day turnout and voter composition is not fully known until Election Day and certainly has the ability to change the outcome of the election.
Demographics of early special election voters helps get a sense of how a race might be shaping up. In particular, party affiliation is usually the best predictor of how someone might vote. Using party affiliation as a proxy for vote choice is trickier in the case of Georgia, however, which does not require voters to self-affiliate with a particular party when registering to vote.
In order to gauge what types of voters might be turning out for the June 20 race, the NBC News Data Analytics Lab analyzed partisan scores computed and provided by TargetSmart. These scores estimate the likelihood that a voter supports the Democratic or Republican Party. Scores closer to 0 represent more likely Republican voters and scores closer to 100 represent more likely Democratic voters.
The figure below, which plots the distribution of partisan scores among April 18 voters overall, shows that Georgia District 6 voters are more Republican-leaning than Democratic.