Ballot box “stuffing” traditionally involves putting extra ballot papers in the ballot box. But in Mozambique it more often means changing the results sheet (“edital”), in many cases adding hundreds of votes.
Details of how we make our estimates are in the 10 November study on http://bit.ly/MozElStuff. The median turnout is around 50% and most polling stations have a turnout between 25% and 75%, with few below 25%. But 7.5% of polling stations (more than 1500) have turnout over 75% and high votes for Nyusi. We consider these to be stuffed ballot boxes, and estimate that they contain 90,000 extra votes for Nyusi.
We also note that 58,000 more people voted from president than voted for parliament (AR) yet no one reports voters putting a ballot paper into the presidential ballot box but not into the parliamentary box. There were, however, many anecdotal reports of individuals putting a few extra ballot papers into the box, often folded together. We take the 58,000 as an estimate of those extra ballot papers. Thus we estimate at least 148,000 stuffed ballot papers for Nyusi.
Inflated registration & ghost voters
Mozambique has 13 parliamentary constituencies. We show that registration was manipulated in 4 of them. In Africa, Gaza and Cabo Delgado the registration was increased to raise the number of votes for presidential candidate Filipe Nyusi, while in Zambézia it was restricted to curb the opposition vote.
AFRICA: Mozambicans in the diaspora can vote for president and elect one seat in parliament each for Africa and Europe. African registration has risen steadily from 44,000 in 2004 to 55,000 in 2009 and 89,000 in 2014. STAE predicted a similar increase for 2019, to 115,000. But the official registration was nearly 213,000. We accept the STAE estimate as plausible but think the huge increase is of fake or “ghost” voters. If we assume the ghosts voted in the same proportion as other voters, they would have given 62,000 votes from Nyusi. We consider these ghost votes as a form of ballot box stuffing for Nyusi.
GAZA: The most discussed scandal of this election has been the over registration in Gaza,
where 1,166,001 people were registered, which is 329,430 more people than the voting age population of Gaza, 836,581, according to the 2017 national population census. Both the head of the National Statistics Institute and the head of the census were dismissed for refusing to bend the census numbers to match the registration.
We note that in nine of the 11 provinces, STAE set a registration target which was 90% of voting age adults, as show by the census. The two exceptions are Gaza, with a target of 137% of voting age adults, and Zambézia discussed below. Nationally, registration was 82.5% of voting age adults.
If we say that the target should be 90% of voting age adults and are generous and say that 95% of them would be registered, then Gaza should only have 715,277 voters. Thus the excess is 450,724 voters – that is 3.4% of the entire national register book. We believe in the accuracy of the 2017 census, which has won praise, so these people cannot exist. Thus we consider them “ghost voters”.
We note that 85% of the ghosts are in the five largest districts, and we note that in four of our five districts (not Chokwe) turnout dropped dramatically compared to previous elections. We attribute this to ghosts not voting, particularly where there were observers. PVT observers noted some polling stations where no one voted except the staff. We believe that of 450,724 ghosts, only 162,000 actually voted, and that they voted for Nyusi.
CABO DELGADO: Three provinces registered more than 95% of their target, while the overall registration was only 91% of the target and 82% of voting age adults. The three top provinces were Gaza and Zambézia at 102% of target and Cabo Delgado at 101%. We do not believe that Cabo Delgado actually registered 91% of voting age adults, beaten only by Gaza at 139%. The third highest percentage registration is Maputo city at 87% of voting age adults. Taking a number between Maputo city and the average of 82%, we say 85% registration could be seen as reasonable for Cabo Delgado, which thus has 6% over registration, about 71,000 ghost voters. We further assume that ghost voters had the same turnout as real voters, 49%, and that they all voted for Nyusi. That gives an extra 35,000 votes for Nyusi.
TOTAL GHOSTS: Adding together the ghost voters in the diaspora (Africa), Gaza and Cabo Delgado gives a massive 259,000 ghost votes for Nyusi.
ZAMBÉZIA: In Zambézia, instead of creating ghost voters, STAE excluded real voters. Zambézia is traditionally an opposition province, and Frelimo was anxious to prevent the election of a charismatic Renamo candidate for governor. In 9 provinces, STAE set a registration target very close to 90% of voting age adults, but uniquely in Zambézia it was set at 77%. This is important because the number of registration brigades assigned was proportional to the target, so Zambézia had fewer registration teams than necessary. There were continuous complaints during the April-May 2019 registration period of registration brigades not being available and, when they were available, the solar panels that charged the laptop computers often did not work. This occurred particularly in strong opposition areas. Despite the problems, there was a strong desire to register, and registration was 102% of the reduced target. Following the reasoning of Cabo Delgado, if we assume that 85% of voters had been able to register in Zambézia, there would have been 186,000 more voters registered. Turnout in Zambézia was 43%, so we assume 80,000 of them would have voted. Because the registration suppression appears to have targeted the strongest opposition areas, we assume 60,000 would have voted for Momade and 20,000 for Nyusi. Therefore we say the under-registration took 40,000 votes from Momade.
AR SEATS: These registration frauds by STAE would also have had an impact on the national parliament (AR, Assembleia da República). Gaza was improperly given 8 extra seats, while Zambézia lost 4 and Nampula, Tete, Manica and Maputo city lost 1 seat each. If we also take away the inflated votes, we estimate that Renamo would have won 5 of the 8 seats (3 in Zambézia, 1 in Nampula, and 1 in Tete). The election law allows the CNE or CC to annul part or all of an election “only if there are illegalities that may substantially influence the general result of the elections”. Surely 5 or 8 parliamentary seats improperly and illegally allocated is a “substantial” change to the result.
Votes taken from opposition
Two kinds of votes are excluded during the counting, blank votes (brancos) in which no candidate is indicated, and invalid votes (nulos) in which it is not clear which candidate has been chosen or where a word has been written on the ballot paper. In the past invalid votes were reconsidered by the CNE. In the 2014 presidential election, polling stations considered 4.4% of ballot papers invalid, but the CNE accepted more than one fourth of those, leaving 3.2% invalid. The reconsideration process was open and it was obvious that two things had occurred in the polling stations. As opposition parties complained repeatedly, opposition ballot papers had been made invalid by a member of the polling station staff adding a fingerprint for a second candidate; this was obvious in a significant number of occurrences where a series of a dozen or more ballot papers would have the same extra fingerprint in exactly the same place on the ballot paper. In 2014 a new fraud was noted in which a group of valid opposition ballot papers had simply been included with invalid votes, probably moved from one pile on the floor to another in the dark during the count late at night in the polling station. There is no longer any check, but we assume no change in the 1.2% of ballot papers improperly included in the nulos. We take 1.2% of the 6.8 million votes to be votes for the opposition improperly taken, about 82,000 votes.
Blank votes are never checked. In this election the total number of blank votes was 4.2% of all votes. Again the distribution of blank votes should follow a normal distribution and we assume any blank votes over 8.5% of the vote in a polling station are improperly taken from the opposition. Using the PVT data, we estimate 28,000 blank votes above 8.5% in individual polling stations. Thus we add these to estimate 110,000 votes improperly taken from the opposition.
