The needle is back. Here’s how it works.
The needle is an innovative forecasting tool that was created by The New York Times and debuted in 2016. It is intended to help you understand what the votes tallied so far suggest about the most likely winner in key contests.
Tonight, it will return for the Georgia runoff.
The needle projects the final result based on an analysis of the vote that’s been counted and an estimate of how many votes are still left to be counted. In states where granular data is available, such as Georgia, it is able to operate especially quickly by analyzing precinct-level returns that are broken out by method of vote.
The needle compares the election results with our pre-election expectations for each county or precinct to determine whether candidates are overperforming or underperforming expectations. It then estimates how the remaining votes will break, based on demographic patterns in the results counted so far.
Much like a weather forecast, the needle cannot be absolutely certain of a result, but the running estimate reflects a range of possibilities that add up to a probability of who will win. For example, if the needle shows that a candidate or a party has a 75 percent chance of succeeding, you should expect that there are scenarios in which the other candidate can win. After all, when the National Weather Service says the chance of rain is 25 percent, you still might bring an umbrella with you when you leave the house.
It’s the smartest way to watch election returns and the best gauge of what could happen on election night.
Our needle starts off with a baseline estimate of how every county or precinct will vote and how the votes will break, whether for the Republican or for the Democrat. These estimates are based on recent election results, historical trends, demographic data and the most recent polling.
Our estimate of how many votes will be cast is based on how many people voted, voted early, requested an absentee or cast a ballot in the last few election cycles. We then adjust to account for other dynamics, like shifts in demographics.
Before election night, the needle starts with these assumptions, meaning there is already a probability for each race. But once the vote tallies start coming in, the needle goes into action, constantly shifting its estimates based on the new data.
For example, if turnout is higher in suburban areas than we had estimated, we will adjust our assumptions for total voters for a similar area. If those areas tend to favor Republicans, that will increase the probability of a Republican win. Similarly, if results show a starker shift toward the Democratic candidates in rural areas than expected, we will adjust the prediction of Democratic votes for similar areas.