EPEC’s tech lead Jon Lareau logged all of the election night return files posted by the VA department of elections (“ELECT”) at five-minute increments on Election Day in order to plot the results over time as the numbers came in later that night.
The data is from this link on ELECT’s website: https://enr.elections.virginia.gov/results/public/api/elections/Virginia/2023-Nov-Gen/files/json
The data feed used a simple “wget” script to grab this file once every five minutes (approximately).
However, when time came to plot the results, EPEC found some data curves that defy explanation. Take, for example, the VA House of Delegates race in the 22nd District, which Lareau plotted straight from the Dept. of Elections feed:
When accumulating counts of ballots, one does not expect vote tallies to go down, let alone oscillate back and forth.
But that’s what we see in the results.
Election technology experts would point out that raw data coming into the main voting database, in this case VERIS, can be filled with errors in the rush to get results out across thousands of voting precincts. That’s fair.
Your humble EPEC volunteers have been a part of that process of transporting the thumb drives with the results of the ballot scans from the precinct to the locality, which then reports out to the state. The goal is to be swift, but accurate to get the thumb drives delivered to the central reporting district in the intense phase that follows the closing of polls. There is a lot to do. We realize errors may happen while the results are being transferred from thumb drives to the central database.
But the way the results fluctuated up and down among Republican and Democrat votes, as if in tandem in some data plots, warrants a closer look. For example here’s how House District 70 race was plotted as the results came in:
This is not the only race where Lareau found ballot curves that have a decrease in one of the ballot counts after a data update. Of the 183 races whose data feed he looked at so far, 79 had a ballot trace that had its count total reduced after a data update.
Here are a few more examples:
Again, one expects some issues and corrections to the election night reporting data. But when 43% of the races sampled have obvious data quality issues like this, we think it deserves some explanation. EPEC has reached out to the Dept. of Elections to ask:
- Why do 79 (and counting) races (~43% of races sampled) in the VA election night reporting have obvious issues where the vote totals decreased after a data reporting update?
- What was the cause?
- Why was it not caught by your QA/QC procedures?
- How will you be addressing it going forward?
Lareau’s full post is here: Fluctuations in election night reporting ballot counts for multiple races in VA 2023 General Election.