Previously I wrote about finding In-Person Early Vote records inserted into the Daily Absentee List (DAL) records after the close of early voting in VA in 2021. Well, theres been quite a bit of activity since then and I have somme updates to share.

I originally discovered this issue and began digging into it around Nov 8th 2021, and finally published it to my blog on Dec 10th 2021. At the same time, queries were sent through the lawyers for the RPV to the ELECT Chairman (Chris Piper) and to a number of registrars to attempt to determine the cause of this issue, but no response was supplied. I also raised this issue to my local Board of Elections chair, and requested that ELECT comment on the matter through their official twitter account.

Since that time I have continued to publish my findings, have continued to request responses from ELECT, and have offered to work with them to address and resolve these discrepancies. I know ELECT pays attention to this site and my twitter account, as they have quietly corrected both their website and data files after I have pointed out other errors and discrepancies. Additionally, Chris Piper has continued to publicly insist that there were no major issues in either the 2020 or 2021 election (including under questioning by the VA Senate P&E Committee), and neither he nor any member of ELECT has publicly acknowledged any of the issues I have raised … besides the aforementioned changing of their site contents, of course. I have thankfully had a few local board of elections members work with me, as well as a few local registrars … but I did not see any meaningful response or engagement from anyone at ELECT until Feb 23rd 2022 as discussed below.

On Feb 22nd 2022, I was invited to participate in a meeting arranged by VA State Senator Amanda Chase with the VA Attorney Generals office to discuss election integrity issues. I specifically cited a number of the issues that I’ve documented here on my blog, including the added DAL entries, as justification for my belief that there is an arguable case to be made for there being criminal gross negligence and maladministration at ELECT with respect to the administration and handling of VA election data.

That meeting apparently shook some things loose. Good.

The day after that meeting Chris Piper finally sent a response at 10:45am to our inquiry on the subject of the added DAL entries. It is quoted below:

While I am glad that he finally responded, his technical reasoning does not address all of the symptoms that are observed:

• He states the cause was due to the EPB’s from one vendor, but the data shows DAL additions being attributed to multiple localities that use different EPB vendors.
• His explanation does not address the distribution of the numbers of DAL additions across all of the precincts observed. A misconfigured or malfunctioning poll book would affect all of the check-ins entered on it, not just a sporadic few.
• This also does not seem like a minor issue as its affecting thousands of voters ballots and voter history. So I’m rather concerned with Mr. Piper’s attitude toward this issue, as well as others. It needs to be addressed as a logic and accuracy testing issue, as a matter of procedures and training, in addition to simply asking vendors to add checks to their software. Also, will this be addressed in VERIS at all, or if/when a replacement system is put in place?
• In response to his last paragraph, I will simply note that I have been actively and consistently working to raise all of these issues through official channels … through official requests via the RPV, working with local election board members and registrars, and asking for input from elect through social media. I have not made accusations of malicious or nefarious intent, but I do think there is plenty of evidence to make the case of incompetence +/or gross negligence with our elections data … which is actually a crime in VA … and Chris Piper has been the head of ELECT and the man responsible for ensuring our election data during that time (he is stepping down effective March 11th).

Since his response was sent, a few additional things have occurred:

(A) The AG’s office informed us that they are actually required by law to treat ELECT as their client and defend them from any accusations of wrongdoing. This is frustrating as there does not seem to be any responsive cognizant authority that is able to act on this matter in the interest of the public. This is not a local jurisdictional issue as it affects voters statewide, and is therefore in the purview of the AG as the Department of Elections has been heretofore dismissive and non-responsive of these matters. I am not a lawyer, however.

(B) I was able to connect with the Loudoun County Deputy Director of Elections (Richard Keech) as well as the Prince William County registrar (Eric Olsen) and have been working through the finer details of Chris’s explanation to verify and validate at the local level. [Note: I previously had Richard erroneously listed here as the Registrar instead of the Deputy Directory]. Both Richard and Eric are continuing to look into the matter, and I continue to work with them to get to the bottom of this issue.

• Richard confirmed his belief that the bad OS system date on Election Day EPBs was responsible for the errors, however with some slight differences in the details from Piper’s description. There were multiple vendors affected, not just one. Per Richard, the problem appeared to be that a number of Loudoun poll-books (regardless of vendor) that were used for Election Day had been in storage so long that their batteries had completely depleted. When they were finally powered up, their OS system clocks had a wildly incorrect date. The hardware used was a mixture of Samsung SM-T720 and iPad tablets, depending on poll-book application vendor. The hardware was purchased separately through local contracts with CDW, and the software was uploaded and configured by the vendors.
• In Loudoun, all of the EPBs went through logic and accuracy testing before the election per Richard, but it does not appear that the procedures for the Logic and Accuracy testing had any specific checks for OS date settings.
• In Prince William County the registrar (Eric) was not aware of any issues with the system clocks on the poll-books, and he was skeptical of the distribution of the small numbers of added DAL entries. He noted, as I did above, that if a poll book was misconfigured it would affect all of the records that passed through it, not just a small handful. He also noticed that there was a discrepancy with the attribution of polling place names that I had extracted from the DAL files, where some of the names did not correspond with actual polling places in PWC. He has stated he will look into the matter and get back to me. I will update my blog when he does so.
• From my communications with Richard, the VERIS system imports a text based file for processing voter credit and does not have any special checks against the dates for in-person vs early voting records. Hence, why this issue can impact multiple vendors if their applications use the system clock to date-stamp their exported txt files for upload into VERIS.

(C) I have reworked my code per my conversations with Ricky and Eric, and fixed a few bugs and parsing errors along the way. Most notably, there are a number of missing or malformed field values in the raw DAL files that were being parsed into ‘undefined’ categorical values by the default csv MATLAB parser. These ‘undefined’ values, even when located in unimportant fields in the row of a MATLAB table, can cause the entire row to be incomparable to other entries when performing logical operations. I have adjusted my parser and logic to account for +/or ignore these entries as necessary. Additionally I had previously looked for new entries by comparing values across entire rows, but have adjusted to now only look at voter ID numbers that have not been seen previously, in order to omit those entries that had simply been adjusted (address changes, etc) after the fact, or that contained ‘undefined’ field elements as mentioned previously. Also I noticed that some of the dal files had duplicate records of Approved and On-Machine records for the same voter ID. While that is an issue in itself, I de-duplicated those entries for this analysis. This new logic gives the updated results presented below, with a total number of discrepancies now at 2820.

I will note that I am still a little skeptical of the “bad date” explanation as being a complete answer to this issue, as it does not adequately explain the distribution of small numbers of discrepancies attributed to multiple precincts, for one thing. While the bad date may explain part of the issue, it does not adequately account for all of the observed effects, IMO. For example, in Loudoun there are 26 precincts listed below that have inserted DAL records attributed to them. Many of these precincts having only 1 or 2 associated records. If the bad OS date explanation is to blame, then (a) there must have been at least 26 poll-books, one at each precinct, in Loudoun with misconfigured and incorrect dates AND (b) many of these poll-books were used to only check in 1 or 2 people total, as anyone checked in on a misconfigured poll-book would have their voter credit/DAL file entries affected. This would have to have been replicated at ALL of the precincts in ALL of the localities listed below. While the above scenario is admittedly possible, I find it rather implausible.

Update 2022-03-20

I’ve heard back from both Ricky Keech (Loudoun) and Eric Olsen (PWC).

Eric looked into the 10 entries for PWC and all of them were Military voters who did actually walk in, in person, to the registrars office and vote on machine absentee after the close of early voting, as allowed for by state law. So everything checks out for PWC.

Per discussion with Ricky, there were two issues: The first being a number of pollbooks for four specific precincts that were used for Election Day having the wrong date setting as discussed above. The other issues was possibly the connectivity issue of the Pollbook to the servers in South Riding during the early voting period that had to be hand corrected. Ricky’s explanation e-mail to me is copied below.

Hi Jon,

Following up on our conversation the other day.  So, I did some more digging and was able to figure out what happened.  The bulk of the voters (1213) were from the four precincts that had a tablet with the wrong date on it.  That accounts for all voters in precincts 214, 416, 628, and 708.  I had a staff person go back and pull out the tablets used in those precincts, and we confirmed that each of those four had one table with the wrong date.

That leaves 141 voters that received ‘credit’ after election day.  Once we had narrowed it down, I looked for patterns and noticed all the remaining precincts were in South Riding.  That jogged my memory and led me to the solution.  When I ran the final reports on the Sunday before the election, I noticed that the number showing as voting early seemed to be off by 100 or so.  This was odd because our daily checked in count and voted count reconciled at every site every day.  So, I went back and compared the voters checked into our pollbooks at early voting to the voters with early voting (On Machine) credit in VERIS and found that there were 137 voters who voted on October 19 at the Dulles South EV site and for some reason did not have credit.  I worked on this Monday to make sure it was right, and it was, none of those voters had credit.  This could either have been a connectivity issue at Dulles South EV site OR an issue with VERIS when the data was uploaded to mark the voters.  I can say definitively that the number checked in on the pollbooks at that site on that day and the number of people who put ballots into the machine was correct, we check that constantly and the observers on site checked as well.  I can guarantee that if there had been a discrepancy, we’d have heard about it right away.

So, after determining that was exactly what happened I uploaded credit for those voters at 2:06:51pm on Wednesday, November 3 and the upload completed processing at 2:07:07pm.

When we spoke the other day, I thought it was likely a connectivity issue, but now I’m not entirely sure that’s the case, as if the connection wasn’t working the numbers should have been off.  And they were correct on the devices at the EV site and my laptop here at the office.  Everything matched.

So long story short, we did an audit, discovered missing credit from one early voting site on one day, and corrected it.

The other four voters were people who voted an emergency early voting ballot on Monday, November 1.

Richard Keech, Deputy Director of Elections for Loudoun County, Mar 11 2022 email to Jon Lareau

Locality	COUNT
LOUDOUN COUNTY	1344
HANOVER COUNTY	1302
CHESAPEAKE CITY	92
PRINCE WILLIAM COUNTY	10
HENRICO COUNTY	7
WINCHESTER CITY	7
CHARLES CITY COUNTY	5
CAMPBELL COUNTY	3
CHARLOTTE COUNTY	3
FAUQUIER COUNTY	3
LUNENBURG COUNTY	3
WASHINGTON COUNTY	3
ALEXANDRIA CITY	2
AMELIA COUNTY	2
AMHERST COUNTY	2
BATH COUNTY	2
CAROLINE COUNTY	2
FALLS CHURCH CITY	2
HENRY COUNTY	2
NORTHAMPTON COUNTY	2
ORANGE COUNTY	2
ROANOKE CITY	2
VIRGINIA BEACH CITY	2
ALLEGHANY COUNTY	1
APPOMATTOX COUNTY	1
BLAND COUNTY	1
CHARLOTTESVILLE CITY	1
CLARKE COUNTY	1
CULPEPER COUNTY	1
ESSEX COUNTY	1
FAIRFAX CITY	1
FLOYD COUNTY	1
KING GEORGE COUNTY	1
MIDDLESEX COUNTY	1
NELSON COUNTY	1
NOTTOWAY COUNTY	1
POQUOSON CITY	1
STAFFORD COUNTY	1
STAUNTON CITY	1

LOCALITY	PRECINCT	COUNT
HANOVER COUNTY	704 – ELMONT	667
HANOVER COUNTY	602 – LEE DAVIS	635
LOUDOUN COUNTY	416 – HAMILTON	443
LOUDOUN COUNTY	214 – SUGARLAND NORTH	344
LOUDOUN COUNTY	708 – SENECA	319
LOUDOUN COUNTY	628 – MOOREFIELD STATION	97
LOUDOUN COUNTY	319 – JOHN CHAMPE	21
LOUDOUN COUNTY	313 – PINEBROOK	16
LOUDOUN COUNTY	112 – FREEDOM	13
LOUDOUN COUNTY	122 – HUTCHISON FARM	11
LOUDOUN COUNTY	126-GOSHEN POST	10
CHESAPEAKE CITY	055 – GEORGETOWN EAST	8
LOUDOUN COUNTY	107 – LITTLE RIVER	8
CHESAPEAKE CITY	053 – FAIRWAYS	7
LOUDOUN COUNTY	121 – TOWN HALL	7
LOUDOUN COUNTY	316 – CREIGHTON’S CORNER	7
LOUDOUN COUNTY	318 – MADISON’S TRUST	7
CHESAPEAKE CITY	008 – SOUTH NORFOLK RECREATION	6
CHESAPEAKE CITY	012 – GEORGETOWN	6
LOUDOUN COUNTY	114 – DULLES SOUTH	6
CHESAPEAKE CITY	018 – INDIAN RIVER	5
LOUDOUN COUNTY	124 – LIBERTY	5
LOUDOUN COUNTY	320 – STONE HILL	5
CHESAPEAKE CITY	029 – TANGLEWOOD	4
CHESAPEAKE CITY	042 – PARKWAYS	4
CHESAPEAKE CITY	059 – CLEARFIELD	4
CHESAPEAKE CITY	065 – WATERWAY II	4
LOUDOUN COUNTY	119 – ARCOLA	4
LOUDOUN COUNTY	322-BUFFALO TRAIL	4
CHESAPEAKE CITY	005 – CRESTWOOD	3
CHESAPEAKE CITY	022 – NORFOLK HIGHLANDS	3
CHESAPEAKE CITY	057 – CYPRESS	3
LOUDOUN COUNTY	120 – LUNSFORD	3
LOUDOUN COUNTY	123 – CARDINAL RIDGE	3
WINCHESTER CITY	101 – MERRIMANS	3
AMHERST COUNTY	501 – MADISON	2
CHARLES CITY COUNTY	101 – PRECINCT 1-1	2
CHARLES CITY COUNTY	301 – PRECINCT 3-1	2
CHARLOTTE COUNTY	702 – BACON/SAXE	2
CHESAPEAKE CITY	010 – OSCAR SMITH	2
CHESAPEAKE CITY	014 – GRASSFIELD	2
CHESAPEAKE CITY	015 – GREENBRIER MIDDLE SCHOOL	2
CHESAPEAKE CITY	016 – HICKORY GROVE	2
CHESAPEAKE CITY	023 – OAK GROVE	2
CHESAPEAKE CITY	024 – OAKLETTE	2
CHESAPEAKE CITY	031 – CARVER SCHOOL	2
CHESAPEAKE CITY	032 – PROVIDENCE	2
CHESAPEAKE CITY	034 – HICKORY MIDDLE SCHOOL	2
CHESAPEAKE CITY	043 – PLEASANT CROSSING	2
CHESAPEAKE CITY	056 – GREEN TREE	2
FALLS CHURCH CITY	003 – THIRD WARD	2
LOUDOUN COUNTY	302 – ROUND HILL	2
LOUDOUN COUNTY	308 – ST LOUIS	2
LOUDOUN COUNTY	309 – ALDIE	2
LOUDOUN COUNTY	617 – OAK GROVE	2
ORANGE COUNTY	101 – ONE WEST	2
PRINCE WILLIAM COUNTY	409 – TYLER	2
PRINCE WILLIAM COUNTY	513 – LYNNWOOD	2
PRINCE WILLIAM COUNTY	712 – LEESYLVANIA	2
WASHINGTON COUNTY	701 – HIGH POINT	2
WINCHESTER CITY	201 – VIRGINIA AVENUE	2
ALEXANDRIA CITY	110 – CHARLES HOUSTON CENTER	1
ALEXANDRIA CITY	201 – NAOMI L. BROOKS SCHOOL	1
ALLEGHANY COUNTY	101 – ARRITT	1
AMELIA COUNTY	301 – NUMBER THREE	1
AMELIA COUNTY	501 – NUMBER FIVE	1
APPOMATTOX COUNTY	401 – COURTHOUSE	1
BATH COUNTY	101 – WARM SPRINGS	1
BATH COUNTY	201 – HOT SPRINGS	1
BLAND COUNTY	301 – HOLLYBROOK	1
CAMPBELL COUNTY	102 – NEW LONDON	1
CAMPBELL COUNTY	402 – COURT HOUSE	1
CAMPBELL COUNTY	602 – CONCORD	1
CAROLINE COUNTY	202 – SOUTH MADISON	1
CAROLINE COUNTY	401 – DAWN	1
CHARLES CITY COUNTY	201 – PRECINCT 2-1	1
CHARLOTTE COUNTY	201 – RED OAK WYLLIESBURG	1
CHARLOTTESVILLE CITY	102 – CLARK	1
CHESAPEAKE CITY	006 – DEEP CREEK	1
CHESAPEAKE CITY	009 – BELLS MILL	1
CHESAPEAKE CITY	011 – GENEVA PARK	1
CHESAPEAKE CITY	020 – E W CHITTUM	1
CHESAPEAKE CITY	033 – WESTOVER	1
CHESAPEAKE CITY	046 – BELLS MILL II	1
CHESAPEAKE CITY	048 – JOLLIFF MIDDLE SCHOOL	1
CHESAPEAKE CITY	049 – WATERWAY	1
CHESAPEAKE CITY	050 – RIVER WALK	1
CHESAPEAKE CITY	051 – COOPERS WAY	1
CHESAPEAKE CITY	062 – FENTRESS	1
CHESAPEAKE CITY	063 – POPLAR BRANCH	1
CHESAPEAKE CITY	064 – DEEP CREEK II	1
CLARKE COUNTY	301 – MILLWOOD	1
CULPEPER COUNTY	303 – CARDOVA	1
ESSEX COUNTY	201 – NORTH	1
FAIRFAX CITY	001 – ONE	1
FAUQUIER COUNTY	202 – AIRLIE	1
FAUQUIER COUNTY	404 – SPRINGS VALLEY	1
FAUQUIER COUNTY	501 – THE PLAINS	1
FLOYD COUNTY	301 – COURTHOUSE	1
HENRICO COUNTY	105 – GREENDALE	1
HENRICO COUNTY	209 – GLEN LEA	1
HENRICO COUNTY	304 – JACKSON DAVIS	1
HENRICO COUNTY	316 – COLONIAL TRAIL	1
HENRICO COUNTY	416 – SPOTTSWOOD	1
HENRICO COUNTY	506 – EANES	1
HENRICO COUNTY	513 – PLEASANTS	1
HENRY COUNTY	203 – HORSEPASTURE #2	1
HENRY COUNTY	501 – BASSETT NUMBER ONE	1
KING GEORGE COUNTY	101 – COURTHOUSE	1
LOUDOUN COUNTY	108 – MERCER	1
LOUDOUN COUNTY	118 – MOOREFIELD	1
LOUDOUN COUNTY	401 – WEST LOVETTSVILLE	1
LUNENBURG COUNTY	301 – ROSEBUD	1
LUNENBURG COUNTY	501 – REEDY CREEK	1
LUNENBURG COUNTY	502 – PEOPLES COMMUNITY CENTER	1
MIDDLESEX COUNTY	501 – WILTON	1
NELSON COUNTY	401 – ROSELAND	1
NORTHAMPTON COUNTY	201 – PRECINCT 2-1	1
NORTHAMPTON COUNTY	401 – PRECINCT 4-1	1
NOTTOWAY COUNTY	201 – PRECINCT 2-1	1
POQUOSON CITY	001 – CENTRAL	1
PRINCE WILLIAM COUNTY	103 – GLENKIRK	1
PRINCE WILLIAM COUNTY	210 – PENN	1
PRINCE WILLIAM COUNTY	305 – PATTIE	1
PRINCE WILLIAM COUNTY	311 – SWANS CREEK	1
ROANOKE CITY	014 – Crystal Spring	1
ROANOKE CITY	019 – Forest Park	1
STAFFORD COUNTY	702 – WHITSON	1
STAUNTON CITY	301 – WARD NO 3	1
VIRGINIA BEACH CITY	030 – RED WING	1
VIRGINIA BEACH CITY	063 – CULVER	1
WASHINGTON COUNTY	302 – SOUTH ABINGDON	1
WINCHESTER CITY	301 – WAR MEMORIAL	1
WINCHESTER CITY	402 – ROLLING HILLS	1

The MATLAB code for generating the above is given below. The raw time-stamped DAL data files, as downloaded from the ELECT website, are loaded from the ‘droot’ directory tree as shown below, and I only utilize the latest daily download of DAL file data for simplicity.

warning off all
% Data directory root
droot = 'SourceData/DAL/2021/';

% Gets the list of DAL files that were downloaded from the ELECT provided
% URL over the course of the 2021 Election.
files = dir([droot,'raw/**/Daily_Absentee_List_*.csv']);
matc = regexp({files.name}, 'Daily_Absentee_List_\d+(T\d+)?.csv','match');
matc = find(~cellfun(@isempty,matc));
files = files(matc);

% Only process the last updated DAL for each day.  I downloaded multiple
% times per day, but we will just take the last file downloaded each day
% for simplicity here.
matc = regexp({files.name}, '(\d+)(T\d+)?','tokens');
fd = []; pc = 0; ic = 0; idx = [];
for i=1:numel(matc)
    if isempty(regexp(matc{i}{1}{1},'^2021.*'))
        fd(i) = datenum(matc{i}{1}{1},'mmddyyyy');
    else
        fd(i) = datenum(matc{i}{1}{1},'yyyymmdd');
    end
    datestr(fd(i))
    if pc ~= fd(i)
        ic = ic+1;
    end
    idx(ic) = i;
    pc = fd(i);
end
files = files(idx);

% Now that we have our list of files, lets process.
seen = [];
firstseen = [];
astats = [];
astatsbc = {};
cumOnMachine = [];
T = [];
for i = 1:numel(files)
    % Extract the date of the ELECT data pull.  Note that the first few
    % days I was running the script I was not including the time of the
    % pull when I was pulling the data fro the ELECT url and writing to
    % disk, so there's soe special logic here to handle that issue.
    matc = regexp(files(i).name, '(\d+)(T\d+)?','tokens');
    matc = [matc{1}{:}];
    if isempty(regexp(matc,'^2021.*'))
        fdn = datenum(matc,'mmddyyyy')+.5;
    else
        fdn= datenum(matc,'yyyymmddTHHMMSS');
    end
    fds = datestr(fdn,30)
    fdt = datetime(fdn,'ConvertFrom','datenum');

    % Move a copy to the 'byDay' folder so we can keep a reference to the
    % data that went into this analysis.
    dal2021filename = [files(i).folder,filesep,files(i).name];
    ofn = [droot,'byDay/Daily_Absentee_List_',fds,'.csv'];
    if ~exist(ofn)
        copyfile(dal2021filename,ofn);
    end

    % Import the DAL file
    dal2021 = import2021DALfile(dal2021filename, [2, Inf]);

    % Cleanup and handle undefined or imssing values.  
    dal2021.CITY(isundefined(dal2021.CITY)) = 'UNDEFINED';
    dal2021.STATE(isundefined(dal2021.STATE)) = 'UNDEFINED';
    dal2021.ZIP(isundefined(dal2021.ZIP)) = 'UNDEFINED';
    dal2021.COUNTRY(isundefined(dal2021.COUNTRY)) = 'UNDEFINED';

    % Do some basic indexing of different DAL status categories and combinations
    appvd = dal2021.APP_STATUS == 'Approved' ;
    aiv2021 = appvd & dal2021.BALLOT_STATUS == 'Issued';
    amv2021 = appvd & dal2021.BALLOT_STATUS == 'Marked';
    aomv2021 = appvd & dal2021.BALLOT_STATUS == 'On Machine';
    appv2021 = appvd & dal2021.BALLOT_STATUS == 'Pre-Processed';
    afwv2021 = appvd & dal2021.BALLOT_STATUS == 'FWAB';
    appmv2021 = amv2021 | aomv2021 | appv2021 | afwv2021; % Approved and Countable

    % Accumulate the stats for each DAL file
    rstats = table(fdt,sum(aiv2021),sum(amv2021),sum(aomv2021),...
        sum(appv2021),sum(afwv2021),sum(appmv2021),'VariableNames',{'DALFileDate',...
        'NumIssued','NumMarked','NumOnMachine','NumPreProcessed','NumFWAB','NumCountable'});
    astats = [astats; rstats];

    % Write out the entries that were approved and countable in this DAL
    % file
    ofn = [droot,'byDayCountable/Daily_Absentee_List_Countable_',fds,'.csv'];
    if ~exist(ofn)
        writetable(dal2021(appmv2021,:),ofn);
    end

    % Write out the entries that were approved, countable and marked as 'On
    % Machine' (i.e. an Early In-Person Voter Check-In) in this DAL file
    ofn = [droot,'byDayCountableOnMachine/Daily_Absentee_List_Countable_OnMachine',fds,'.csv'];
    if ~exist(ofn)
        writetable(dal2021(aomv2021,:),ofn);
    end

    % Since the DAL file grows over time, we're going to try and figure out
    % which On-Machine entries in each new file:
    %       (a) We've seen before and are still listed.
    %       (b) We haven't seen before (a NEW entry).
    %       (c) We've seen before but the listing is missing (a DELETED
    %           entry).
    if isempty(T)
        % Only applicable for the first file we process.
        T = dal2021(aomv2021,:);

        % There are sometimes duplicate uid numbers in the approved and
        % on-machine counts!  This is a problem in and of itself, but not
        % the problem I'm trying to focus on at the moment.  So I'm going
        % to remove any duplicated rows based on UID.
        [uid, ia, ib] = unique(T.identification_number);
        T = T(ia,:);
        inew = (1:size(T,1))';
        ideleted = [];
        firstseen = repmat(fdt,size(T,1),1);
    else
        dOM = dal2021(aomv2021,:);

        % There are sometimes duplicate uid numbers in the approved and
        % on-machine counts!  This is a problem in and of itself, but not
        % the problem I'm trying to focus on at the moment.  So I'm going
        % to remove any duplicated rows based on UID.
        [uid, ia, ib] = unique(dOM.identification_number);
        dOM = dOM(ia,:);

        %[~,ileft,iright] = innerjoin(T,dOM);
        [~,ileft,iright] = intersect(T.identification_number,dOM.identification_number);

        % 'inew' will be a boolean vector representing those entries in
        % 'dOM' that are new On-Machine records
        inew = true(size(dOM,1),1);
        inew(iright) = false;

        % 'ideleted' will be a boolean vector representing those entries in
        % 'T' that are missing On-Machine records in 'dOM'
        ideleted = true(size(T,1),1);
        ideleted(ileft) = false;
        T = [T;dOM(inew,:)];
        firstseen = [firstseen; repmat(fdt,sum(inew),1)];
    end
    clf;
    plot(astats.DALFileDate,astats{:,2:end},'LineWidth',2);
    grid on;
    grid minor;
    legend(astats.Properties.VariableNames(2:end),'Location','NorthWest');
    xlabel('Date of DAL file pull from ELECT');
    ylabel('Counts');
    drawnow;
end
ofn = [droot,'byDayStats.csv'];
writetable(astats,ofn);
T.FirstSeen = firstseen;
ofn = [droot,'onMachineRecords.csv'];
writetable(T,ofn);
ofn = [droot,'onMachineRecords_missing.csv'];
writetable(T(ideleted,:),ofn);
cutoffDate = datetime('2021-11-01');
ofn = [droot,'onMachineRecords_after_20211101.csv'];
writetable(T(T.FirstSeen >= cutoffDate,:),ofn);
Ta = T(T.FirstSeen >= cutoffDate,:);
[ulocality,ia,ib] = unique(Ta.LOCALITY_NAME);
clocality = accumarray(ib,1,size(ulocality));
Tu = table(ulocality,clocality,'VariableNames',{'Locality','COUNT'});
ofn = [droot,'numOnMachineRecords_after_20211101_byLocality.csv'];
writetable(Tu,ofn);
Ta = T(T.FirstSeen >= cutoffDate,:);
[uprecinct,ia,ib] = unique(join([string(Ta.LOCALITY_NAME),string(Ta.PRECINCT_NAME)]));
cprecinct = accumarray(ib,1,size(uprecinct));
Tu = table(Ta.LOCALITY_NAME(ia),Ta.PRECINCT_NAME(ia),cprecinct,'VariableNames',{'LOCALITY','PRECINCT','COUNT'});
ofn = [droot,'numOnMachineRecords_after_20211101_byLocalityByPrecinct.csv'];
writetable(Tu,ofn);

The adjusted MATLAB parser function is listed below:

function dal = import2021DALfile(filename, dataLines)
%IMPORTFILE Import data from a text file
%  DAILYABSENTEELIST10162021 = IMPORTFILE(FILENAME) reads data from text
%  file FILENAME for the default selection.  Returns the data as a table.
%
%  DAILYABSENTEELIST10162021 = IMPORTFILE(FILE, DATALINES) reads data
%  for the specified row interval(s) of text file FILENAME. Specify
%  DATALINES as a positive scalar integer or a N-by-2 array of positive
%  scalar integers for dis-contiguous row intervals.
%
%  Example:
%  dal = import2021DALfile("SourceData/DAL/Daily_Absentee_List_10162021.csv", [2, Inf]);
%
%  See also READTABLE.
%
% Auto-generated by MATLAB on 16-Oct-2021 14:19:26
%% Input handling
% If dataLines is not specified, define defaults
if nargin < 2
    dataLines = [2, Inf];
end
%% Set up the Import Options and import the data
opts = delimitedTextImportOptions("NumVariables", 38);

% Specify range and delimiter
opts.DataLines = dataLines;
opts.Delimiter = ",";

% Specify column names and types
opts.VariableNames = ["ELECTION_NAME", "LOCALITY_CODE", "LOCALITY_NAME", "PRECINCT_CODE", "PRECINCT_NAME", "LAST_NAME", "FIRST_NAME", "MIDDLE_NAME", "SUFFIX", "ADDRESS_LINE_1", "ADDRESS_LINE_2", "ADDRESS_LINE_3", "CITY", "STATE", "ZIP", "COUNTRY", "INTERNATIONAL", "EMAIL_ADDRESS", "FAX", "VOTER_TYPE", "ONGOING", "APP_RECIEPT_DATE", "APP_STATUS", "BALLOT_RECEIPT_DATE", "BALLOT_STATUS", "identification_number", "PROTECTED", "CONG_CODE_VALUE", "STSENATE_CODE_VALUE", "STHOUSE_CODE_VALUE", "AB_ADDRESS_LINE_1", "AB_ADDRESS_LINE_2", "AB_ADDRESS_LINE_3", "AB_CITY", "AB_STATE", "AB_ZIP", "BALLOTSTATUSREASON", "Ballot_Comment"];
opts.VariableTypes = ["string", "string", "categorical", "string", "string", "string", "string", "string", "string", "string", "string", "string", "categorical", "categorical", "categorical", "categorical", "categorical", "string", "string", "string", "categorical", "string", "categorical", "string", "categorical", "double", "string", "categorical", "categorical", "categorical", "string", "string", "string", "string", "string", "string", "string", "string"];

% Specify file level properties
opts.ExtraColumnsRule = "ignore";
opts.EmptyLineRule = "read";

% Specify variable properties
opts = setvaropts(opts, ["ELECTION_NAME", "LOCALITY_CODE", "PRECINCT_CODE", "PRECINCT_NAME", "LAST_NAME", "FIRST_NAME", "MIDDLE_NAME", "SUFFIX", "ADDRESS_LINE_1", "ADDRESS_LINE_2", "ADDRESS_LINE_3", "EMAIL_ADDRESS", "FAX", "VOTER_TYPE", "APP_RECIEPT_DATE", "BALLOT_RECEIPT_DATE", "PROTECTED", "AB_ADDRESS_LINE_1", "AB_ADDRESS_LINE_2", "AB_ADDRESS_LINE_3", "AB_CITY", "BALLOTSTATUSREASON", "Ballot_Comment"], "WhitespaceRule", "preserve");
opts = setvaropts(opts, ["ELECTION_NAME", "LOCALITY_CODE", "LOCALITY_NAME", "PRECINCT_CODE", "PRECINCT_NAME", "LAST_NAME", "FIRST_NAME", "MIDDLE_NAME", "SUFFIX", "ADDRESS_LINE_1", "ADDRESS_LINE_2", "ADDRESS_LINE_3", "CITY", "STATE", "COUNTRY", "INTERNATIONAL", "EMAIL_ADDRESS", "FAX", "VOTER_TYPE", "ONGOING", "APP_RECIEPT_DATE", "APP_STATUS", "BALLOT_RECEIPT_DATE", "BALLOT_STATUS", "PROTECTED", "CONG_CODE_VALUE", "STSENATE_CODE_VALUE", "STHOUSE_CODE_VALUE", "AB_ADDRESS_LINE_1", "AB_ADDRESS_LINE_2", "AB_ADDRESS_LINE_3", "AB_CITY", "AB_STATE", "BALLOTSTATUSREASON", "Ballot_Comment"], "EmptyFieldRule", "auto");

% Import the data
dal = readtable(filename, opts);

% Perform some cleanup on commonly found issues...
dal.Ballot_Comment = strrep(dal.Ballot_Comment,char([13,10]),". ");
dal.BALLOTSTATUSREASON = strrep(dal.BALLOTSTATUSREASON,char([13,10]),". ");
dal.LOCALITY_NAME = categorical(strtrim(string(dal.LOCALITY_NAME)));
dal.PRECINCT_NAME = categorical(regexprep(strtrim(string(dal.PRECINCT_NAME)),'^(\d+)( +)(\w+)','$1 - $3'));
end

On a spur of curiosity I went back to some of the data provided by the VA dept of elections (“ELECT”) for both the 2020 and 2021 elections and ran a new data consistency test …

I have a copy of the final Daily Absentee List (DAL) for both 2020 and 2021. I also have a copy of the paired Registered Voter List (RVL) and Voter History List (VHL) generated shortly after the close of the 2021 General Election and within a few moments of each other.

I was curious what the percentage of approved and counted absentee ballots from the DAL is that do NOT have an associated “voter credit” in the VHL for both 2020 and 2021. If ELECT’s data is accurate the number should be ideally 0, but most official thresholds for acceptability that I’ve seen for accuracy in election data systems hover somewhere around 0.1%. (0.1% is a fairly consistent standard that I’ve seen per the documentation for various localities Risk Limiting Audits, and the Election Scanner Certification procedures, etc.) The VHL should cover all of the activity for the last four years, but to ensure that I’m accounting for people that might have been officially removed from the RVL and VHL since the 2020 election (due to death, moving out of state, etc), I only run this test on the subset of the entries in the DAL that still have a valid listings in the RVL.

The results are below. Both years seem to have a high amount of discrepancies compared to the 0.1% threshold, with 2020’s discrepancy percentage being over 3x the percentage computed for 2021.

Year	Percent of Counted DAL Ballots without Voter Credit
2020	1.352%
2021	0.449%

For those interested in the computation, the MATLAB pseudo-code is given below. I can’t actually link to the source data files because of VA’s draconian restrictions on redistributing the contents of the DAL, RVL and VHL data files.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% We first compute the indices of the DAL entries that represent 
% approved and countable ballots ...
%
% 'dal2020' and 'dal2021' variables are the imported DAL tables 
% 'VAVoteHistory' is the imported Voter History List
% 'RegisteredVoterList' is the Registered Voter List
% 
% All four of the above are imported directly from the CSV 
% files provided from the VA Department of elections with 
% very little error checking save for obvious whitespace or 
% line ending checks, etc.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

aiv2021 = dal2021.APP_STATUS == 'Approved' & dal2021.BALLOT_STATUS == 'Issued';
amv2021 = dal2021.APP_STATUS == 'Approved' & dal2021.BALLOT_STATUS == 'Marked';
aomv2021 = dal2021.APP_STATUS == 'Approved' & dal2021.BALLOT_STATUS == 'On Machine';
appv2021 = dal2021.APP_STATUS == 'Approved' & dal2021.BALLOT_STATUS == 'Pre-Processed';
afwv2021 = dal2021.APP_STATUS == 'Approved' & dal2021.BALLOT_STATUS == 'FWAB';
counted2021 = amv2021 | aomv2021 | appv2021 | afwv2021; % Approved and Countable
    
aiv2020 = dal2020.APP_STATUS == 'Approved' & dal2020.BALLOT_STATUS == 'Issued';
amv2020 = dal2020.APP_STATUS == 'Approved' & dal2020.BALLOT_STATUS == 'Marked';
aomv2020 = dal2020.APP_STATUS == 'Approved' & dal2020.BALLOT_STATUS == 'On Machine';
appv2020 = dal2020.APP_STATUS == 'Approved' & dal2020.BALLOT_STATUS == 'Pre-Processed';
afwv2020 = dal2020.APP_STATUS == 'Approved' & dal2020.BALLOT_STATUS == 'FWAB';
counted2020 = amv2020 | aomv2020 | appv2020 | afwv2020; % Approved and Countable

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Next we compute the indices in the VHL that represent 
% 2020 and 2021 General Election entries for voter credit
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
valid_2020_entries = strcmpi(strtrim(string(VAVoteHistory.ELECTION_NAME)), '2020 November General');
valid_2021_entries = strcmpi(strtrim(string(VAVoteHistory.ELECTION_NAME)), '2021 November General');


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% We use the MATLAB intersect function to make sure that 
% we are only using DAL entries that are still in the RVL 
% and therefore are possible to be present in the VHL and 
% compute the percentages.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
[did,iida,iidb] = intersect(dal2020.identification_number(counted2020), ...
    RegisteredVoterList.IDENTIFICATION_NUMBER);
[vid,iida,iidb] = intersect(VAVoteHistory.IDENTIFICATION_NUMBER(valid_2020_entries),...
    RegisteredVoterList.IDENTIFICATION_NUMBER);
[iid,iida,iidb] = intersect(did,vid);
pct2020 = (1-numel(iida) / numel(did)) * 100

[did,iida,iidb] = intersect(dal2021.identification_number(counted2021), ...
    RegisteredVoterList.IDENTIFICATION_NUMBER);
[vid,iida,iidb] = intersect(VAVoteHistory.IDENTIFICATION_NUMBER(valid_2021_entries),...
    RegisteredVoterList.IDENTIFICATION_NUMBER);
[iid,iida,iidb] = intersect(did,vid);
pct2021 = (1-numel(iida) / numel(did)) * 100

Originally Posted 01-31-2022.

As of 03-20-2022, the latest version of my master spreadsheet posted below.

This includes the updates from my discussions with the York County registrar (see here), the comparison to DAL file entries (here), and the latest updates and comparisons to Statement of Results (SOR) / Machine Tape records, including ABLEMARLE, ACCOMACK COUNTY (partial), FAIRFAX COUNTY, FREDERICK COUNTY, GOOCHLAND COUNTY, HENRICO COUNTY (partial), ISLE OF WIGHT, LOUDOUN COUNTY (partial), LOUISA COUNTY, NORFOLK CITY, ORANGE COUNTY, POWHATAN COUNTY (partial), PRINCE WILLIAM COUNTY, STAFFORD COUNTY, SPOTSYLVANIA COUNTY, SUFFOLK COUNTY and YORK COUNTY. There’s more to process, but I wanted to put out a status update.

The previous result between the “official” CSV results and the “official” DAL file results has not changed at 5,766 Net discrepancies and 17,194 absolute. Latest update includes some cleaning up of the spreadsheet formulas and making sure that I am only accounting for discrepancies for those rows where we actually have verified transcriptions and/or tape images. The current discrepancies in the records between the CSV and the SOR / Machine Tapes is 4,549 Net discrepancies and 27,905 Absolute discrepancies. “Net” discrepancies are summed such that +/- deltas per candidate can cancel out, whereas “Absolute” discrepancies are summed using the absolute value. I have only been focusing on the Governors race at the moment and not any of the down-ballot races. Again many many “thank you’s” to the volunteers at VFA who have been helping to put all of this together.

The more I dig, the more I am disturbed by the VA Department of Elections (a.k.a. “ELECT”) lack of data integrity and “data hygiene”. If any private citizen or company was as reckless and error prone with financial or other legal records, they’d be in jail and fined out of existence. And I’m being very very very generous as I am assuming “innocent” incompetence on the part of our elected officials and ELECT. (Although I’m not exactly sure which is worse … discovering that there is a hypothetical cabal of bad actors responsible behind the curtain, or that the vast majority of our elected officials are simply just incompetent, and the system as-is is unreliable.)

Many have been asking me for updates and status as to the various datasets I’ve been working on. Process is slow going, as I do this in my spare time, but progress IS being made. (With specific thanks to all of my peeps over at VFA on telegram who have stepped up to the plate in collecting machine tape transcriptions and images.)

I’ve previously posted on basic stats and other findings (here, here, here, here, here).

I have been working through all of the data and combined inputs from:

• the “2021 November General.csv” file, dated 2021-12-11_15:58:46
• the “Turnout-2021 November General.csv”, dated 12-12-2021
• the Daily Absentee List (DAL) File, downloaded on 2021-12-05_11:22:04
• images of the machine tapes, as available for verification and validation.

There is a whole lot of work left to do to digitize and transcribe the machine tapes and compare them to the results of the data as presented by ELECT. But even ignoring the machine tapes for the moment, we can see from the attached master spreadsheet that there are a NET 5,766 and ABSOLUTE 17,194 ballot accounting discrepancies between the various ELECT provided datasets for the 2021 Election.

Columns L,M,N,O,P are derived from the “2021 November General.csv” file as of  12-11-2020 provided by VA ELECT on their website

Columns S-Y are derived from the “Turnout-2021 November General.csv” file as of 12-12-2021 provided by VA ELECT on their website

Columns Z-AP are derived from the Daily Absentee List (DAL) file provided by VA ELECT, downloaded 2021-12-05T11-22-04

Columns H-K are generated from in person inspection of the machine tapes for each precinct

Columns E-G are the differences between the machine tapes and the sums reported in the (2021 November Genreal.csv” file

Column Q is the difference between the CAP precinct data reported in “2021 November General.csv” and the corresponding “Turnout-2021 November General.csv” or DAL derived data. For Rows with precinct == “## Provisional” : Q is the difference between column P and the sum of all of the column S for the corresponding precincts. For Rows with precinct == “## AB – Central Absentee Precinct” : Q is equal to column AM minus column P for both that row and the corresponding row of Post-Election-Day (PE) counts For Rows with precinct == “## EV – Central Absentee Precinct” : Q is equal to column AN minus column P

Column R is the absolute value of column Q

Summing column Q (the net discrepancy) gives 5,766

Summing column R (the absolute discrepancy) gives 17,194

I stress again that this is based on data directly supplied by the VA Department of Elections. This is not based on estimates or approximations. There should be Zero (as in Zip, Zilch, Nada) discrepancies between their own distinct datasets that are supposed to be representing the same real world event (the election).

By law in VA copies of the return sheets and machine tapes, or Statement of Results (SOR), from tabulation machines are public record and must be made available for 60 days after the election. (see VA 24.2-658) A grassroots team of a number of volunteers (coordinated through Virginia First Audits) made a concerted effort to collect images of a number of these tapes for specific localities of interest. I have been going through these tapes and comparing the results they show with the official results presented by the VA Department of Elections. I started with Prince William County, which is where I live.

The reported totals of counted votes on the tapes should match exactly to what is reported as counted votes by the VA Department of Elections, after correctly omitting over-votes/under-votes and separating out provisional ballots (which is simple to do with the way VA reported its data). I have confirmed this should be the case via multiple email and phone conversations with election experts and officials, in an effort to make sure that I wasn’t missing some obvious process or fact that could account for the discrepancies.

An example of the information included on these tapes is shown below in Figure 1. The exposed section of the far left tape shows the vote totals for that machine, per candidate, summarized by House and Congressional district combination. The exposed section of the middle tape in the image shows the report summary of the total ballots counted for each House and Congressional district combination. The exposed section of the far right tape top in the image shows the signatures of the election officials on the tape, and the bottom show the introductory information at the beginning of the report that identifies the machine information. Each machine is dedicated to a specific precinct and ballot type and are labeled as such.

Computation of Results

The machine tape images corresponding to absentee ballot categories (“ABS” – Mail-In Absentee, “EARLY” – In Person Early Voting, and “PE” – Mail-In ballots received Post Election but with valid postmark) were examined by hand and the tabulated results for each precinct were extracted. These results were then compared with the reported totals from the VA Department of Elections website dated 12-11-2021 (https://apps.elections.virginia.gov/SBE_CSV/ELECTIONS/ELECTIONRESULTS/2021/2021%20November%20General%20.csv).

The “Margin” column in the tables below accounts for individual candidate totals that could not be read from the tape images due to imaging errors, but the end of tape reports listed the total number of ballots processed for that precinct so a maximum limit (“Margin”) of the number of votes is known. We are currently in the process of going back to the registrar and filling in the missing data.

All of the tables below correctly omit undervotes/overvotes and provisional ballot counts. (The machine tapes do not include provisional ballots.)

The results presented below in Table 1, Table 2 and Table 3 show a consistent pattern of the machine tapes and the Department of Elections data giving nearly identical matches to the In-Person Early vote, but different results for the

Prince William County – Congressional District 01

	Democrat	Libertarian	Republican	Total	Margin
TAPE ABS	5076	89	1842	7028	21
TAPE EARLY	9748	64	7639	17451	0
TAPE PE	193	9	86	341	53
CSV ABS	5155	90	1884	7129
CSV EARLY	9746	64	7638	17448
CSV PE	251	13	96	360
PWC CD 01	Est. Delta Dem	Est. Delta Lib	Est. Delta Rep	Est. Delta Tot	Est. Delta Tot – Margin
CSV ABS – TAPE ABS	79	1	42	122	101 (~ 1.42 % diff)
CSV EARLY – TAPE EARLY	-2	0	-1	-3	-3 (~ -.02 % diff)
CSV PE – TAPE PE	58	4	10	72	19 (~ 5.28 % diff)

Prince William County, Congressional District 10

	Democrat	Libertarian	Republican	Total	Margin
TAPE ABS	2292	28	996	3316	0
TAPE EARLY	5809	28	6045	11882	0
TAPE PE	57	4	24	119	34
CSV ABS	2320	28	1009	3357
CSV EARLY	5809	28	6045	11882
CSV PE	79	5	43	127
PWC CD 10	Est. Delta Dem	Est. Delta Lib	Est. Delta Rep	Est. Delta Tot	Est. Delta Tot – Inc Slop
CSV ABS – TAPE ABS	28	0	13	41	41 (~ 1.22 % diff)
CSV EARLY – TAPE EARLY	0	0	0	0	0 (~ 0.00 % diff)
CSV PE – TAPE PE	22	1	19	42	8 (~ 6.30 % diff)

Prince William County, Congressional District 11

	Democrat	Libertarian	Republican	Total	Margin
TAPE ABS	5307	93	1333	6733	0
TAPE EARLY	9780	60	3578	13418	0
TAPE PE	152	6	36	293	99
CSV ABS	5375	93	1314	6782
CSV EARLY	9780	60	3578	13418
CSV PE	237	10	63	310
PWC CD 11	Est. Delta Dem	Est. Delta Lib	Est. Delta Rep	Est. Delta Tot	Est. Delta Tot – Inc Slop
CSV ABS – TAPE ABS	68	0	-19	49	49 (~ 0.72 % diff)
CSV EARLY – TAPE EARLY	0	0	0	0	0 (~ 0.00 % diff)
CSV PE – TAPE PE	85	4	27	116	17 (~ 5.48 % diff)

Source Data:

Raw Machine Tape Images

2021 November General – 2021-12-11T15:58:46