All of Mozambique’s elections have had some fraud and sloppiness, but it was more extreme in the 2019 general elections. For example, there have always been reports of small numbers of ballot papers pre-marked for Frelimo in the hands of voters or Frelimo officials outside of polling stations – which requires the collusion of polling station staff who must illegally remove ballots from books in polling stations or at STAEs.
In 2019 these reports were much more common and there were more reports during the counts of three to five ballot papers folded together and deposited in the ballot box, which had to be separated during the count and thus obviously had been put into the box together.
Most fraud in the polling stations is carried out by, or with the collusion of, polling station staff. This occurs at three points in the process. Most polling stations are in classrooms and when the votes are counted one-by-one, tick-marks are put on the blackboard, grouped in 5s. But there may be a hundred or more groups of 5, and it is easy to “miscount” – indeed, the official STAE manual for polling station staff blatantly shows how to do this: http://bit.ly/STAE-board
Ballot papers are placed in piles on the floor. Many polling stations have no electricity, so it is easy in the dark to move ballot papers from the opposition pile to the blank or invalid pile. In previous elections invalid votes were checked by national CNE, in public, and there were often handfuls of valid votes in the invalid pile which were accepted as valid and added by the CNE to opposition vote totals. There was also ample evidence of opposition votes being invalidated by adding a fingerprint for a second candidate. Probably because this was being picked up, increasing numbers of opposition votes were put in the blank pile, which has never been checked. Rules were changed for 2019, and invalid votes are no longer checked, meaning that significant numbers of valid votes remain in the invalid piles.
The official results sheets (editais) for the three elections – president, parliament, and provincial assembly – are only written when all of the counting of all three is finished. This is often at the early hours of the morning by tired staff with party delegates and observers sleeping or having gone home. Few observers or delegates remember what was written on the blackboard or may simply allow a count of the piles of ballots to check. Thus staff are free to inflate and invent numbers.
Ballot box stuffing tends to occur in areas which already strongly support Frelimo, because polling staff already back Frelimo and will not object to adding Frelimo votes. The most gross examples are in Gaza and Tete, and are the same districts that have seen large scale ballot box stuffing in past elections. CC data shows that in Gaza five entire districts each voted 99% for Nyusi, and had unbelievably high turnouts: Massangena andn Chicualacuala 96%; Chigubo 95%, Mabalane 92% and Mapai 91% turnout. And the same two Tete districts turn up in each election: Zumbu 90% turnout and 89% Frelimo, and Changara 86% turnout but 97% Frelimo. These are not single polling stations, but entire districts.
Whereas ballot box stuffing occurs in predominantly Frelimo zones, taking votes away from the opposition is more common in contested areas and particularly in the north, notably Niassa, Cabo Delgado, Nampula and Zambézia, where there is less ballot box stuffing. Tete is a divided province and has districts in both categories.
For the provincial assemblies (APs) with its short ballot paper, we find it bizarre and impossible that 36 of 164 entire districts had more than 8.5% blank votes. Cabo Delgado had the highest levels of blank votes, the most extreme being Namuno with 18% blank votes, Chiure district with 17% blank and Mecufi 16%. Cabo Delgado had become a highly contested province, because Renamo unexpectedly won Chiure town in 2018 municipal elections. With only 4 parties on the AP ballot paper, were 1 in 6 voters really unable to choose, or were votes taken away from Renamo?
The PVT sample data for the presidential vote showed similar extremes for invalid votes in individual polling stations in the presidential election. A polling station at Monequera primary school, Ulongué, Angonia district, in Tete had 45% invalid votes. There were polling stations with more than 25% invalid votes in Tete (especially Angonia and Maravia), Nampula (Mossuril), Niassa (Mecanhelas), Zambézia (Molumbo), and Manica.We do not believe that up to half of voters made mistakes.
We cannot identify small fraud, a few extra ballot papers or minor changes to the edital. But we can use statistical techniques to identify large scale fraud – exceptionally high turnout implies large scale ballot box stuffing, very high invalid or blank votes suggest votes taken from the opposition, and registering more people than there are voting age adults seems impossible. In a study published 10 November 2019, the bulletin estimated large-scale fraud and gave details of our methods: http://bit.ly/MozElStuff. We summarise this below.
Some of the data we use here comes from an extrapolation of the PVT (parallel vote tabulation) which covered 2,507 of the 20,162 polling station in a randomly selected sample of 12.4%.
A key to any statistical analysis is that electoral data should have what is called a “normal distribution”, which means it that most points are near the mean and are equally distributed on the two sides of the mean. Scatter diagrams are useful because they give a visual representation of data. On our charts, the vertical axis is the percentage of the vote for Nyusi and the horizontal axis is the turnout. The points are PVT polling stations, so we can “see” the polling stations. Thus our polling stations should be concentrated in the centre of the scatter diagram. We show here scatter diagrams of the two provinces, Maputo City and Tete.
Maputo shows a largely normal distribution – turnout was 57% and Nuysi had 69% of the vote – both in the centre of the blob of polling stations.
Now look at Tete. Polling stations are more scattered but the turnout of 60% and the vote for Nyusi of 77% are in the middle of the polling station points. However, note the large number of polling stations in the upper right corner, with a disproportionate number over 80% for Nyusi and far too many also with over 75% turnout. This is not a statistically “normal” distribution. Looking at the diagram, we can “see” the polling stations on the upper right with improper and excessive turnout.
Using statistical techniques we conclude that the three types of fraud – ballot box stuffing, spoiling ballot papers to take votes from the opposition, and over and under registration – inflated Nyusi’s margin of victory by at least 557,000 fraudulent votes.
