The Numbers Didn't Add Up
Armenia held its parliamentary election on Sunday, June 7, 2026. They are still recounting the votes. Here is why.
Preamble
Back in 2014, when I was running for re-election as a school board trustee in Toronto’s municipal election, a different ward of the city faced a mini election scandal. In Ward 19 (Scarborough Centre), challenger Scott Harrison was neck-and-neck against incumbent school board trustee David Smith. Late into the night, the Toronto Elections website showed that, with 100% of precincts* reporting, Harrison had squeaked out ahead and won the election. Some time later, however—I can’t remember if it was the same night or the next morning—the numbers changed. They still showed 100% of precincts reporting, but Smith’s tally had increased, showing that he had won with 10,405 votes compared to Harrison’s 10,266. According to the rules, if a candidate requested a recount, they had to pay the election staff salaries involved from their own funds. Harrison came to the Board and asked us to fund the recount, given the circumstances. A representative of Toronto Elections also came to that Board meeting to offer an explanation: The electronic tabulators used by Toronto Elections send each precinct’s totals to the central office through a mobile data connection. In one specific precinct, the tabulator had sent in a total of 0 votes for Smith (an unlikely scenario, since he was the incumbent). Upon double-checking the numbers (which also get called in and reported on a paper protocol signed by the local precinct supervisor), Toronto Elections staff corrected the error and entered the number of votes for Smith in that precinct. It just so happened that it made the difference between him winning the election or not.
This is a blog about electoral reform in Armenia, not Toronto, but I share the story to highlight that things can go wrong in counting and adding votes… even in Canada, even with multi-million dollar budgets and sophisticated equipment. For this reason, it is important to have observers watching the count, signing the protocol, and then checking the official results posted on the website to confirm that the correct numbers are being reported and summed.
The Total Number of Votes Should Equal the Sum of the Votes from Each Precinct
Armenia is learning a similar lesson this week. Observers matter, and correcting seemingly small errors can make a difference in who gets to sit as a representative. Armenia’s Central Electoral Commission (CEC) made some glaring mistakes in tabulating the results of the 2,006 precincts. Note that there are 2,005 physical precinct voting stations within the country; an addition 537 voters have the ability to vote over the internet. These are mostly Armenian diplomats and their families posted abroad.
The initial precinct-by-precinct results that the CEC posted on its website (a Microsoft Excel file) included a totalization row at the bottom. However, the formulas in that row only added up the votes from the 2,005 physical precinct voting stations and not the votes cast electronically. (The electronic votes were included in a row in the Excel file, just not added up.) Separate from the published Excel file, a scanned protocol, signed by the seven members of the CEC reports the official results (prior to any recounts). This signed PDF file actually double-counted the votes cast electronically, i.e. those votes were added twice.
You could download those files for yourself and see that the math didn’t add up. However, the CEC website, www.elections.am, has been blocking internet users from outside Armenia from accessing the site since the polls closed on election night. Consequently, archiving websites like the Wayback Machine (based in the US) are not capturing snapshots in time of the official election site and the files uploaded to it. So here they are:
The very first verification that should be done is that the total vote count should add up the individual vote counts from each precinct accurately. When this basic step gets bungled, whether intentionally or not, it opens an opportunity for the credibility of the election to be chipped away at.
The head of the CEC, Vahagn Hovakimyan, announced that one political party, the Prosperous Armenia Party (PAP) led by Gagik Tsarukyan, fell just short of the 4% threshold, announcing that they had received 3.996% of the vote. But this number is based on the double-counted electronic votes. Anyone doing the math for themselves would get 3.997%; they were 41 votes shy of the threshold.
The Votes Listed in the Spreadsheet For a Precinct Should Match the Official Record Signed at the Precinct
The second verification that should be done is that each individual precinct result should be the same as the numbers that were counted together by the precinct electoral commission (PEC) members and documented in the precinct protocol. Here, again, irregularities were reported. The PAP, realizing they were inches away from the 4% threshold, complained that the numbers on the signed precinct protocols did not match the numbers in the CEC’s precinct-by-precinct results Excel file. For example:
At Precinct 27/7 (Aragyugh village in Kotayk region), the PEC protocol says that PAP received 77 votes, but the CEC website gives them only 1 vote in that precinct (shortfall of 76 votes).
At Precinct 12/20 (Aygestan village near Ararat region’s regional capital Artashat), the PEC protocol says that PAP received 26 votes, but the CEC website gives them only 1 vote in that precinct (shortfall of 25 votes).
At Precinct 27/26 (Charentsavan city in Kotayk region), the PEC protocol says that PAP received 19 votes, but the CEC website gives them only 3 votes in that precinct (shortfall of 16 votes).
At Precinct 38/53 (Varagavan village near Berd in Tavush region), the PEC protocol says that PAP received 7 votes, but the CEC website gives them only 4 votes in that precinct (shortfall of 3 votes).
Those four precincts alone, coming to a shortfall of 120 votes, would seemingly make the difference between the PAP receiving 5 MPs or 0 MPs, and consequently whether Nikol Pashinyan’s Civil Contract Party would have more or less than a three-fifths supermajority threshold needed to pass certain changes such as amendments to the Electoral Code.
Now it’s one thing to claim an irregularity and another to have it substantiated. After follow-up checks, though, eventually CEC Chair Vahagn Hovakimyan accepted that there had been inputting errors in five precincts.
The “Official” Inaccuracy Number
It is often said that, to be eligible for seats in the new parliament, a political party needs to have attained 4% of the votes cast (the electoral threshold). In most cases, that explanation is good enough. However, Armenia’s Electoral Code actually sets the threshold at 4% of “the sum of the total number of ballot papers with affirmative vote [i.e. not invalid] and the number of inaccuracies”. So what is this “number of inaccuracies”?
There is a column in the results for “inaccuracies”. This number has two components. The first component involves the holographic stamps that are placed on the corner of a ballot paper. Each precinct is allocated a certain number of stamps. At the end of voting, the number of unused stamps left plus the number of voters that walked in should equal the number of stamps that precinct had at the start of the day. Any difference is recorded as the “first inaccuracy”.
The second component is that the total number of voters that walked in (and signed their name on the voter list) should be the same as the number of valid and invalid votes in that precinct. If there are more votes than voters, the difference between them is recorded as the “second inaccuracy”.
The first inaccuracy and the second inaccuracy are added and recorded as that precinct’s “official inaccuracy”.
Now, in an ideal world, the official inaccuracy should be zero for all precincts. What never made sense to me is that a party on the cusp of passing the electoral threshold is penalized for these inaccuracies. Since these official inaccuracies are added to the number of valid votes for the purposes of determining what the 4% threshold will be, the higher the inaccuracies, the more votes an upstart party would need to get to qualify for the distribution of MP seats.
Pending the outcome of the recounts, according to the CEC’s preliminary results, the total number of inaccuracies was 1279. Therefore, a party will need an extra 52 votes than they would have if the Electoral Code did not include these amounts. If the Electoral Code did not have this provision, even before correcting the “inputting errors”, the PAP would have received 4.00071% of the valid votes cast and would have passed the 4% threshold. For the record, back when I was at the table on electoral reform efforts in Armenia, I had recommended removing consideration of these inaccuracies in the electoral threshold calculation, but wasn’t successful in convincing the decision-makers.
As a last point on these inaccuracies, I’ll add that it would be worth checking out why there were:
34 more votes than voters in Precinct 2/17
194 fewer votes than voters in Precinct 3/26 (14.5% of the votes from that precinct are unaccounted for—not invalid, unaccounted for)
34 fewer votes than voters in Precinct 5/7
90 more votes than voters in Precinct 7/12
300 fewer votes than voters in Precinct 14/6 (26.4% of the voters there)
26 more votes than voters in Precinct 21/1
41 more votes than voters in Precinct 23/6
96 fewer votes than voters in Precinct 27/2
100 more votes than voters in Precinct 34/29
Even the electronic voting results had one more vote cast (411) than the number of voters recorded (410). How does that happen? To boot, in this case, the “official inaccuracy” for the electronic voting should technically be 1, but the CEC’s preliminary result spreadsheet left that cell blank.
Preventing Perceived Conflict of Interest
It is important to state that these irregularities are being examined and a recount is currently underway. In a tense situation like this, it doesn’t help when the impartiality of the head of the CEC is under doubt. Prior to being appointed as Chair of the CEC, Vahagn Hovakimyan was a Member of Parliament for the Civil Contract Party. Even the circumstances of his appointment itself were controversial, which you can read about by scrolling to the end of this article I wrote in 2023.
Let’s say that, after all the recounts are done, it really does turn out that the PAP didn’t meet the 4% threshold and the Civil Contract Party did achieve a three-fifths supermajority. Who would be believe him? It is for this reason that it is important for an election administration body to be free from even the perception of conflict of interest.
To make the situation smell even worse, Gagik Tsarukyan was arrested by law enforcement while the current recounts are ongoing, despite a requirement in Armenian law for the CEC to consent to any such arrest of a candidate during an election period (which seemingly was never given).
There will be more to comment on about this election, but it didn’t start off on the right foot.
*In Canada, precincts are usually called polls. I used the term precinct throughout this article to avoid confusion with survey polls.