Some have suggested that voter suppression may at least partially explain how often we see a higher fraction of votes for the Republican candidate in precincts where more ballots were cast.
The idea is that if there are more obstacles to voting in heavily Democratic precincts fewer voters will cast ballots in these precincts. This would have two effects:
- Making these precincts appear smaller, since fewer ballots were cast
- Reducing the apparent turnout at these precincts, thus making it seem as though Democrats are apathetic
There are a number of ways that voters can be suppressed. Broken voting machines or other technical difficulties as were well documented in Detroit in 2016, for instance.
Another way voters of one party might be suppressed is by manipulation of voter registration data. This can be done overtly via systems like CrossCheck or by systematic purges that disenfranchise classes of voters, like those we have seen recently in Georgia and North Dakota.
Then there is the danger of covert or malicious manipulation of voter registration data by third parties that have gained access to online voter registration databases. A detailed look at several ways online voter registration could be compromised and the possible effects of such a compromise is here.
With this in mind, we decided to look to the data itself. Specifically, examined individual sets of statewide voter registration data for oddities such as very old or underage voters and and multiple voters with the same name in the same household. We also compared sets of voter registration records from different dates in order to see exactly which voters were removed from and added to each data set over a given time period. We also compared data sets to see which records showed changes to voter name, address, political party, and other identifying information including gender and birth date.
We have completed a thorough analysis of 2016 and 2017 registration data from Pennsylvania. We have also looked at the large number of active voters who were removed from Los Angeles County rolls in the weeks before the June 2018 primary, when a “printing error” was blamed for the fact that over 100,000 voters were missing from the registration lists. We are in the process of repeating this analysis on current publicly available registration data from Florida, Ohio, New York, North Carolina, and Washington State.
We will be adding more information to these pages as we complete and document our analyses.
Based on this work, we have two strong recommendations.
- Every state should create a system to monitor changes in voter registration records and to alert voters when any change appears to have been made to his or her record.
- Every state should create a system that allows poll workers and the public to report problems with the registration data at individual precincts.
Voter registration data analyses:
Pennsylvania (2016 – 2017)
California (April 2018)
North Carolina (Fall 2018)
New York (Fall 2018)