Summary:

• Issue #1: There are 99 records of ballots cast, according to the VA Department of Elections (ELECT) Daily Absentee List (DAL) data file that do not have corresponding voter ID listed in Registered Voter List (RVL) data.
• Issue #2: There are 380 records of ballots cast in the DAL where the corresponding RVL record has been listed as “Inactive” since June-30-2023 and no modification to the RVL record has taken place.
• Issue #3: There are 18 records of ballots cast in the DAL where the corresponding RVL record is listed as “Inactive” as of Dec-01-2023, but there has been previous modifications to the RVL record since June-30-2023.
• We are currently reaching out in attempts to contact the VA AG’s office and to provide them the details of this analysis in order to have these anomalies further investigated.

Data files utilized for this analysis:

Our 501c3 EPEC purchased and downloaded the full statewide VA RVL on June-30-2023 from ELECT. We additionally purchased the Monthly Update Service (MUS) package from ELECT, where on the 1st of each month we are provided a list of all of the changes that have occurred to the RVL in the previous month. By applying these changes to our baseline data file, we are able to update our copy of the RVL to reflect the latest state as per ELECT. We can also create a cumulative record of all entries associated with a particular voter ID by simply concatenating all of these datafiles.

Additionally, during the VA 2023 General Election, we purchased access to the Daily Absentee List (DAL) file generated by ELECT that documents all of the transactions associated with early mail-in or in-person voting during the 45 day early voting period. The DAL file we utilized for this analysis was downloaded from ELECT on Nov-13-2023 at 6am EST.

Identification of ballots cast via the DAL file can be performed by checking for rows of the DAL data table that have the APP_STATUS field set to “Approved” and have the BALLOT_STATUS field set to any of the following: “Marked” | “Pre-Processed” | “On Machine” | “FWAB” | “Provisional”.

Once cast ballots are identified in the DAL, the Voter Identification Number can be used to lookup all of the corresponding records in our cumulative RVL data. The data issues summarized above can be directly observed using this process. Due to VA law, we cannot publish the full specific records here in this blog but have summarized, captured and described our process and results.

For Issue #1: If there does not exist a corresponding registration record for cast ballots, then the voter should not have been able to have their mail-in ballot approved or issued, or been able to check-in to an early voting precinct to vote on-machine. If the voter record actually does exists, then why is it not reflected in the data that we purchased from ELECT. Note that all provisional and Same Day Registration (SDR) ballots were required to be entered into the states database (“VERIS”) by the Friday after the election (Fri Nov-10-2023). We specifically waited until we received the Dec-01-2023 MUS data update from ELECT to attempt to perform this or similar analysis in order to ensure that we would not be missing any last minute registrations or RVL updates.

For Issue #2: There are 380 records of ballots being cast in the DAL where the baseline June-30-2023 RVL data file shows the registrant as inactive, and there has been no modifications or adjustments to the record presented in the MUS data files. Therefore these registrants should still have been listed as “Inactive” during the early voting period which started in September through Election Day (Nov 7).

For issue #3: There are 18 records that show the cast ballot is from a registrant that is currently listed as “Inactive” but there has been adjustments to the registration record over the last 6 months. An example of such is below. Note that I have captured the MUS data file generation dates in the 5th column to note when the file was generated and received by us.

In the example given below, the first invalidation operation on the registration record appears in the MUS file dated Sep-01, with the earliest transaction date listed as Aug-29-2023. The ballot application was not received until Sept 26 according to the DAL, so the application should never have been approved or the ballot issued as the registrant status should have been “Invalid” according to the states own data.

(Also … yes … I know there is a typo in the spelling of “APP_RECIEPT_DATE” in the tables below … but this is the data as it comes from ELECT).

APP_RECIEPT_DATE	APP_STATUS	BALLOT_RECEIPT_DATE	BALLOT_STATUS
“2023-09-26 00:00:00”	Approved	“2023-10-19 00:00:00”	Pre-Processed

TRANSACTIONDATE	TRANSACTIONTIME	Trans_Type	NVRAReasonCode	File Source
30-June-2023	12:12:00	BASELINE	BASELINE	Baseline RVL
28-Jul-2023	09:34:03	MODIFY	Change Out	MUS 08/01/2023
28-Jul-2023	09:34:04	MODIFY	Address Change	MUS 08/01/2023
28-Jul-2023	09:34:04	MODIFY	Change In	MUS 09/01/2023
28-Jul-2023	09:34:03	MODIFY	Change Out	MUS 09/01/2023
28-Jul-2023	09:34:04	MODIFY	Address Change	MUS 09/01/2023
28-Jul-2023	09:34:04	MODIFY	Change In	MUS 09/01/2023
29-Aug-2023	11:55:49	MODIFY	Inactivate	MUS 09/01/2023
29-Aug-2023	11:55:49	MODIFY	Inactivate	MUS 10/01/2023

(Updates in red text as of 01-27-2022)

Walt Lantham, the registrar of York county (yes … it’s York County … don’t worry … I asked Walt if he was ok with me identifying him and the locality publicly before making this update) generously reached out and contacted me to try and clear up some of the original numbers I posted below. We had a long and detailed couple of phone calls (over two hours in total) and went over all over these results in detail. It turns out that I was missing the hand counted ballot tally sheets from the data that I had received. This cleared up most of the technical issues with the 2021 data and all but 1 issue with the data below for 2020. The remaining issue is the discrepancy with the official State CSV files, which is outside of Walt’s ability to diagnose. Our discussion also brought up some of the interesting procedural issues and challenges related to data reporting, retention, maintenance and general practices as well. I’ll touch on that in a bit. It was a great conversation and a big thank you to Walt and his staff for taking the time to address these issues in detail.

Special thanks again to the grassroots folks in VFA, and throughout the state, for all their hard work in collecting the machine tape information for various localities. There were a few localities that provided not only the 2021 machine tapes, but the 2020 versions as well! The first one of those that I have compiled all the data for is reported below. As with all of my results, I invite those that wish to replicate, validate or refute to contact me and I will happily provide the necessary details and data files.

I believe we have some significant issues with our elections in VA. The data below helps clearly make that case, in conjunction with the multiple other data analyses that I’ve done and posted here on my blog. The data below is only from a single locality in the state (I am working on others). I am purposefully summarizing the data and abstracting which locality this represents, as I do not want to jeopardize any potential legal proceedings or official investigations. But I do feel strongly that the “fact of” this information needs to be made public and become part of the discussion on the matter. (And … just to be clear … the background image for this blog post is the tapes from a different locality in 2021 only, which I discussed here)

I’ll note that the machine tapes and Statements of Results (SORs) are the “gold standard” to measure the published election results against in VA. Once a vote is cast on a machine, there is no process for it to be removed from the official vote count. Each ballot category is supposed to be scanned into a set of dedicated machines. Provisional ballots have their own dedicated machines that they get scanned into. Mail-In ballots have their own dedicated machines. Early In-Person ballots get their own machines, as do the In-Person Early and Day-Of ballots.

Ballots that cannot be scanned for some reason must be hand counted. This hand counting and verification is performed at the post-election canvass. In my discussion with Walt, we noted the fact that there is no requirement for the hand-count tally worksheets used during the canvass to be included in the statement of results reports that are available for the public to view. There is also no mechanism for the state to track or delineate the numbers of hand-counted ballots, or way for voters to know the various reasons why ballots needed to be hand counted in the first place. But luckily for me … Walt keeps good notes and records, and e-mailed me a copy of the Tally sheets that he preserved from the canvass for my records and to fill in the data gaps.

There should be zero difference from the sum of these reports (including the hand counts) to the reported official results, save for very very very small and rare documented errors and issues. Maaaaaybe 1 or 2 differences here and there, but nowhere near the results presented below when compared to the State CSV files. I have talked to multiple elections officers, public officials, registrars and lawyers and ALL of them have agreed with this sentiment.

Input Data:

The official results are supplied by the VA Dept. of Elections (ELECT) website in a csv file for 2020 (here) and 2021 (here). Additional versions of the official results are reported through a web portal to the ELECT historical archive for 2020 (here). Turnout numbers are supplied similarly from csv files on the ELECT servers for 2020 (here) and 2021 (here). Note that the official CSV and WEB tally results for 2020 show significant differences, yet both are supposed to reflect the “official” results. That, on its own, is a significant issue that warrants further inspection. I’ve computed the deviations between both data sets for this locality, so it’s dealers choice as to which dataset you want to compare against. For 2021 the CSV results are the only ones available.

SOR and machine tape information was obtained by volunteers across the state going to their local registrar or county clerks office. For some counties complete datasets of both 2021 and 2020 election results were obtained, with hand transcriptions of tape returns as well as photographic copies of the originals. The results reported below are for a moderately sized York County and were derived from both hand transcribed results and photographic copy of the machine tapes for both 2020 and 2021. The Statement of Results (SOR) reports and change logs from the county do not document any issues that would account for the discrepancies below. There were no issues presented at the canvass (where the tapes and SOR’s are checked against the state database entries for errors) that would explain the results below that I am aware of. As noted above per my discussions with Walt, and with him providing the copies of the hand-count tally worksheets, we were able to address most of the discrepancies that I came across except for the discrepancy of the 2020 data with the CSV version of the official results posted by VA ELECT. In fact, Walt’s corrections made the discrepancy with the CSV results slightly worse, even though they fixed the discrepancies with the WEB based portal results.

Results:

The tables below show the computed differences of the VA Dept of Elections data minus the collected machine tape results, stratified by major candidate party. The 2020 results are for office of president, the 2021 results for Governors race. The 2021 and 2020 Web results were strictly positive differences, so there is no distinction between the Sum of the differences (the NET change) and the Sum of the Absolute Value of the differences (how many votes were affected). Whereas, the 2020 CSV results had some negative differences, so there are two different ways to compute the effects of these deviations.

The 2021 Governor’s race showed a smaller amount of % deviation from the machine tapes than the 2020 Presidential race for this locality, but still show non-zero differences. (Remember, there should be zero difference.)

It is notable to mention that in 2021 the reporting was changed such that absentee mail-in and early-vote totals are separated in the reported data. The 2021 In-Person Day-Of results were identical to the reported results presented by the VA Department of Elections, with the In-Person Early vote only showing a single vote discrepancy. Almost all of the differences are manifested in the Mail-In or Provisional ballots for the 2021 data.

Likewise for 2020 there is very little difference (a single vote again) in the In-Person Day-Of precinct totals from the reported results. Almost all of the differences can be attributed to Central Absentee Precinct (CAP) data when compared against the state’s official CSV report.

2021 Results

Year	PRECINCT	VDOE-TAPE difference for Democratic candidate	VDOE-TAPE difference for Libertarian candidate	VDOE-TAPE difference for Republican candidate
2021	Total Differences Provisional	22 0	0	18 0
2021	Total Differences AB – Central Absentee Precinct	25 0	0	9 0
2021	Total Differences EV – Central Absentee Precinct	0	0	1
2021	Total Differences PE – Central Absentee Precinct	1 0	0	2 0
2021	Total Differences Day-Of Precincts	0	0	0
	Total Differences	48 0	0	30 1
	Machine Tape Totals	12142 12190	150	17455 17484
	% of Machine Tape Totals	48/12142*100 = 0.40% 0/12190*100 = 0.00%	0/150*100 = 0.00%	30/17455*100 = 0.17% 1/17484*100 = 0.006%

2020 Results (ELECT Web Portal Archive)

Year	PRECINCT	VDOE-TAPE difference for Democratic candidate	VDOE-TAPE difference for Libertarian candidate	VDOE-TAPE difference for Republican candidate
2020	Total Differences AB – Central Absentee Precinct	234 0	10	112 0
2020	Total Differences Provisional	2 0	0	1 0
2020	Total Differences Day-Of Precincts	0	0	1
	Total Differences	236 0	10 0	114 1
	Machine Tape Totals	17447 17683	670 680	20027 20240
	% of Machine Tape Totals	236/17447*100 = 1.35% 0/17447*100 = 0.00%	10/670*100 = 1.49% 0/670*100 = 0.00%	114/20027*100 = 0.57% 1/20027*100 = 0.005%

2020 Results (ELECT CSV Report)

Year	PRECINCT	VDOE-TAPE difference for Democratic candidate	VDOE-TAPE difference for Libertarian candidate	VDOE-TAPE difference for Republican candidate
2020	Total Differences AB – Central Absentee Precinct	77 157	-1 11	11 101
2020	Total Differences Provisional	-114 116	-1 1	-52 53
2020	Total Differences Day-Of Precincts	0	0	1
	Total Differences	-37 273	–2 12	64 155
	Machine Tape Totals	17447 17683	670 680	20027 20240
	% of Machine Tape Total	-37/17447*100 = -0.21% 273/17683*100 = 1.54%	-2/670*100 = -0.30% 12/680*100 = 1.80%	64/20027*100 = 0.32% 155/20240*100 = 0.77%
	Abs Val Total Differences	191 273	2 12	164 155
	Abs Val % of Machine Tape Totals	191/17447*100 = 1.09% 273/17683*100 = 1.54%	2/670*100 = 0.30% 12/680*100 = 1.80%	164/20027*100 = 0.82% 155/20240*100 = 0.77%