New York's "double votes"

It has taken three days to get something out of the new databases, but here it is. The table above shows what looks like double voting happening in New York, as reported in the state’s official voter roll database. I’ve checked 5 snapshots of the database, dated from 10/2021 to 8/2025. For this article, I will call them SS1-SS5 (Snapshot 1 through 5).

In the earliest snapshot, SS1, double votes appear as 2 separately recorded votes, as in this example from the 2020 General (presidential) Election:

2020 GENERAL ELECTION;General Election 2020;20181106 GE;20180913 PR;20161108 GE;Presidential Primary Election 2016;General Election 2012;General Election 2011

All subsequent snapshots of the database that I’ve seen, have added a parenthetical notation of the vote method. In the example below, (P) stands for a vote cast in person at the polling place and an (E) stands for early voting. So this record is telling us that the same voter voted early and then voted again on Election Day.

GENERAL ELECTION(P);2020 GENERAL ELECTION(E);2020 GENERAL ELECTION(P);2019 GENERAL ELECTION(P);2018 GENERAL ELECTION(P)

An oddity of these records is that there weren’t very many 2020 duplicate votes at first, only 108 in S1, but they quickly jumped to very high numbers. SS2 has 225,515 such records. From there, the number continues to increase, until 2025 when it reaches 233,319. This is odd because all of the snapshots postdate the 2020 election by at least a year. These numbers should be stable, but they aren’t.

A change occurred sometime in early 2022. Two new fields were added to the database, “NonStdAddr” which accounted for records with blank addresses found by NYCA (though they still didn’t have addresses), and “AptType”. A third change is that vote method was added in parentheses after the election description. After this was done, or at the same time, the number of double votes skyrockets.

One possible explanation was that when voters moved from one NY county to another, their voter history was transferred in a way that caused it to be double counted. That would explain why voter histories for old elections were retroactively modified.

Picture it this way: the last time you vote is in 2020. Then you move in 2021. In 2022, the database is changed in a way that requires checking the old record for the vote method. Once checked, the vote method is added to the voter history, but the way it is done causes the most recent vote to be imported twice. In this way, a double vote not present in an earlier snapshot suddenly appears in a later one, though both snapshots were made much later than the election.

At first, the records supported the theory. The first 5 records in a row matched the pattern: the SBOEID occurred in two different records that had different county ID numbers. The registration dates were different, and the earlier registration date record had the most recent vote doubled in the other record. That is, of the pair, the double vote only appeared in the more recent record. Also, in every case, there was a name change. In several cases it was clearly a woman changing to a married name, but in one, it was a first name gender change. However, the pattern didn’t hold. For the explanation to work, it had to apply to a very high majority of the records with just a few outliers, but that wasn’t the case.

The spike of double votes in 2020 was also puzzling. Why that year? The number of double votes in 2020 dwarfs all other years, though it was less than most until SS2. For reference, here is the table again:

Notice that the quantity of double votes that suddenly appear in SS2 is close to the numbers in subsequent snapshots, though they gradually increase for almost every year. That makes sense if there was a single event, like the addition of vote method, that caused an initial spike as all records were updated, then a gradual increase over time as people moved in natural patterns.

If the spike was due to a single modifications that affected all previous years, why are they all for the same recent election? If the number of people who moved in state during that time span is roughly consistent, it is inconsistent for every other year.

The face value interpretation is that these are records of fraudulent votes. First, a vote is cast in early or absentee voting, then later, another is cast for the same person at a polling place. How that could happen is hard to imagine, because it would require a total breakdown of the system that uses the voter rolls to record votes specifically to prevent double voting. We hear about people who are stopped from voting in NY for this exact reason. Did the system break down in some places but not others?

The answer is that it doesn’t matter. There are two possibilities:

1. The data is true. If so, these are illegal fraudulent votes that violate every election law and the Constitution.

2. The data is false. If so, the mere existence of these records violates HAVA requirements for accurate records.

I’m not sure which option is right, but don’t think there are any others, and both are bad for the state.

The best “innocent” explanation I can think of is that there was a database modification in 2022 that somehow caused this error and it was never fixed. The problem with that explanation is that, while plausible, it doesn’t completely fit the known data.

As for a nefarious explanation, I can picture the county recording a vote initially, sending it to the state, then overwriting it with another vote later and sending that to the state also, where the original vote isn’t overwritten, because this field is concatenated. However, if that was done, I don’t understand what criminal advantage is obtained by adding double votes years after the fact, primarily to the 2020 election. I also don’t see a mechanism that is logistically feasible without raising red flags even the most incompetent employee would notice.

Bottom line, the data is false or fraudulent. That alone is reason to start an investigation.