When To Expect Catalist’s Post-Election Data

Since publishing this blog post and spreadsheet, there’s been a lot of interest about our post-election data for 2018. We just wanted to post a quick update regarding timing. We are not planning on publishing estimates on election night, but we will start publishing estimates by the end of the week.

The bottleneck here is precinct results. As mentioned in the previous post, we aren’t just conducting another poll. We are going to collect precinct results state-by-state, match them to our database, and then run a series of models to produce the most accurate estimates that we can, based on all of the data.

Unfortunately, most states don’t make the precinct data available immediately. And if it’s a good night for Democrats, there may still be many races too close to call on election night or even the morning after. Many states won’t make the final and official data available for weeks or months after the election. Here’s what our process will be:

1. First, we start with a set of states and congressional districts that have data available immediately. We will collect data and publish results as soon as possible, probably beginning on Friday 11/9, maybe even earlier.

2. Results will also be available across the country at the congressional district and county level. We’ll compare trends that we’re seeing in the granular precinct results, to what we can see at these larger geographies.

3. On or before the weekend, we’ll publish our national estimates, along with anything we learned from the previous steps. At this point, our results will be labeled “preliminary,” and we’ll continue updating them as more precinct data comes in.

4. Eventually, as actual vote history data comes in from the Secretaries of State across the country, we will publish final data.

Depending on your expectations, this might be a huge disappointment or a huge victory. This will certainly be slower than the traditional exit poll, which will be available the night of the election. But for people who are used to waiting months for accurate, validated turnout data, a turnaround time of a few days is a huge leap forward. Apologies for any initial confusion in the last post, hopefully, this clears things up!