Correction Update

St. Louis County Heat Maps

We’re committed to the truth and getting the facts right, which sometimes means correcting errors.

Big thanks to everyone who reached out to us to flag that the St. Louis County heatmap shared on the David Pakman Show earlier this week looked a little off! We dug into the data and found that an exploratory version of the chart had been shared in error, which pulled from a spreadsheet we were still in the process of de-aggregating to the precinct level and accounting for population size, as is standard methodological practice. The original and corrected heatmaps are shown below for comparison.

To clarify:

1. The visual difference in average vote-share between hand-count to machine-count precincts is not especially pronounced when viewed as heat maps.

2. Only 30% of the county’s precincts hand-counted instead of the misstated 50%

3. Candidate Harris won the county overall with 55% of the vote.

The updated St. Louis County data does revise some of the initial observations we shared on the Pakman show. However, the corrected results continue to reveal statistical concerns in the machine count results, in which vote distributions for machine count precincts appear anomalous while hand count precincts appear less so. We expect to publish our full report on this topic in the near future.

Inaccurate Heat Map Comparison:

Updated Heat Map Comparison:

What’s Different?
With the corrected St. Louis County hand-count heatmap, the voteshare comparison between counting types is still visually apparent but less stark.

Another useful approach to visualizing the difference is comparing the distribution of votes across turnout, which also reveals a slight noticeable shift at higher turnout benefiting the Republican Presidential candidate in the machine counting data.

Because both datasets fall closer to the 50% margin we can compare them to a normal distribution.

Hand Count Distribution by Candidate:

When analyzing the hand-count precincts in St. Louis county, we see that voteshare relative to turnout exhibits a more normal behavior for both candidates. We do not see an aggressive shift or change at higher or lower turnout thresholds such as indicated as possible election fraud as seen in Russian elections.

Machine Count Distribution by Candidate:

When analyzing the comparable range of machine-count precincts, there is a slight rightward skew for the Republican candidate. The skew is minor but does demonstrate a slight increase in voteshare in precincts with higher turnout, which can be an indicator of potential vote manipulation.

Summary:

• Hand-count precincts for St. Louis County at the Presidential level seem to exhibit more normal, expected voting behavior

• This is in contrast to the concerns with machine-count precincts with slightly abnormal voteshare versus turnout distributions, and a noticeable shift benefiting the Republican candidate at higher turnout.

• These concerns warrant further investigation into the impacted precincts by evaluating historical trends and wider datasets.

Thank you again for reaching out to us to flag the issue and issue this correction!”