Center for Electoral Forensics

Kagame for Three

[Flag of Rwanda]Rwanda is best known in the West for 100 days in 1994. The Rwandan Genocide saw the deaths of as many as 1,000,000 Rwandans and the migration of as many as three time more. From the ashes of this great disaster, Paul Kagame (a Tutsi) rose up to restore order and bring some semblance of peace to this beautiful land. That was 1994. On August 4, Kagame stood for reelection yet again.

Officially, he received more than 98% of the votes cast. Did we detect issues with this election?

[paul kagame]

Rwandan President Paul Kagame at the ITU Plenipotentiary Meeting in Busan, Korea. Photograph courtesy Veni Markovski.

Rwanda blasted into American awareness during its notorious genocide in 1994, which saw the Hutu majority rise up and attack the Tutsi minority. During the colonial period, Rwanda was ruled by Germany (until the end of World War I) and Belgium (until 1962). Both countries made the Tutsis the leaders of the colony, much to the distaste of the majority Hutu.

At independence in 1962, the Hutu were able to install a president who reflected their majority status. This, in conjunction with its history, sent the country into fits of ethnic violence where the Tutsis tended to attack from neighboring countries and Hutus responded by attacking Tutsis.

In 1972, neighboring Burundi, governed by Tutsis, saw their first apparent genocide. Hutus rose up against the ruling Tutsis and slaughtered them. In retaliation, the ruling Tutsis attacked and killed up to 300,000 Hutu (Totten et al., Century of Genocide).

The Genocide

This resurgent Tutsi strength in Burundi actually led to Hutu Juvénal Habyarimana taking control of the Rwandan government through a military coup. Initially, Habyarimana sought to create policies that were blind to the Hutu-Tutsi chasm. However, as the years passed, Habyarimana became more pro-Hutu and more authoritarian. This led to Tutsis fleeing the country and some Hutus wishing for an end to the ethnic programs.

In 1994, while returning from a regional summit in Dar es Salaam, Tanzania, Habyarimana's plane was shot down while on final approach to Kigali International Airport. The attack was immediately blamed on Tutsi rebels by Hutus in the media. Thus started 100 days of genocide in Rwanda. They were ended only when the Tutsi rebel force made it to Kigali.

At the head of that force was Paul Kagame.


As the head of the army that ended the genocide, Paul Kagame became the de facto leader of Rwanda. This changed in 2003, when Rwanda held its first presidential elections since the genocide. Kagame won with 95% of the votes. In 2010, after his first seven-year term, Kagame stood for reelection. He won that with 93% of the votes.

According to the Rwandan constitution of 2003, the president serves a seven-year term and may be reelected once. Article 101 is very clear on that:

The President of the Republic is elected for a term of seven years renewable only once.

Under no circumstances shall a person hold the office of President of Republic for more than two terms.

However, this article was altered by a referendum in 2015. That referendum changed the constitution to allow a third term for presidents. It also changed "the length of a term from seven to five years and maintain a two-term limit, the rules will not take effect until 2024." Would that mean Kagame could stand for election in 2024 and 2029?

The 2017 Election

Because the 2015 constitutional referendum allowed Kagame to stand for reelection again, he did. In 2003, he received 95% of the votes cast; in 2010, 93%. Both of those elections saw violence and accusations of fraud. In both elections, Kagame denied issues. For instance, after the 2010 election, he stated

Speaking to reporters after voting, Kagame said the process was "very democratic" and dismissed allegations the real opposition was de facto excluded from the vote.

"The people of Rwanda were free to stand for election - those who wanted to - and to qualify, so I see no problem," he said.

"Some sections of the media seem to be reading from a different page."

Thus, in the run-up to the 2017 election, nobody had a doubt who would win (Kagame: "A mere formality"). The only question was whether Kagame would receive more or less than 95% of the vote. The election cycle saw the usual repression of opposition and violence. Election day saw the usual claims of fraud and unfairness. It also saw the Rwandan Supreme Court declare Paul Kagame the winner with over 98% of the vote. This map shows his level of support across the country.

[kagame support]

Support map for Paul Kagame in the 2017 Rwandan presidential election.


It is here that we step in with our analysis. The first thing to check is whether or not the official results of the election seem reasonable. Since Kagame routinely received more than 90% of the votes cast, and since the 2015 referendum passed with more than 98% of the vote, it is reasonable that he actually received this number of votes. Pre-election polling would also help us determine if the results were reasonable, but these were not done is a reliable manner.

Differential Invalidation

Beyond the "smell-test," we can test for differential invalidation. This is a condition where votes for one candidate are invalidated at a higher rate than votes for another candidate. If this exists, it indicates an unfair election.

For an election such as this, where the reported candidate support rates are extreme, the performance of the test is unknown. Here is the invalidation plot for the invalidation rate against the Kagame-support level in the akarere (province).

[invalidation plot all]

Invalidation plot using all provinces (akarere) of Rwanda. Note that the effect appears slight if anything.

Note that the slope is shallow. Also note that the analysis was not able to detect a statistically significant relationship between the two variables. That is, we cannot conclude that there is evidence for differential invalidation (or ballot box stuffing).

Note, however, that there are three provinces that have abnormally high invalidation rates. As they are outliers, it could be that the "rules of invalidation" differ for them. If so, then removing them would allow us to better understand the "normal" rules for the rest of the the country. Here is the invalidation plot with those removed.

[invalidation plot - restricted]

Invalidation plot removing the three outlier provinces (akarere) of Rwanda. Note that the effect remains slight.

Note that the effect remains slight. Here, because the variation in the residuals was reduced by removing the outliers, the effect is now statistically significant. It is still very slight, however. The p-value is only 0.0199. While this is technically less than α = 0.05, I am not comfortable claiming evidence of fraud based solely on this. It is barely significant.


There is a newly proposed method for detecting direct fraud, fraud due to people manually adjusting the natural vote counts. This method was introduced by Arturas Rozenas in his Political Analysis article titled "Detecting Election Fraud from Irregularities in Vote-Share Distributions," called the RKD method (implemented in the R statistical environment as the spikes package).

Because Rwanda publishes the number of registered voters (abagombaga gutora), the number of votes cast (abatoye), and the number of votes for Paul Kagame, we can use this new and exciting method. Even more exciting is that the RKD method allows us to estimate the number of fraudulent divisions, divisions where fraud occurred.

According to this test, the estimated percentage of fraudulent precincts is zero.


And so, we are left with an interesting case. While there is direct evidence of unfairness in this election (candidate intimidation, election violence, suppression of dissent, undermining opposition campaigns, etc.), we were not able to detect evidence of unfairness beyond that.

Remember that electoral forensics is not intended to replace election observers. It merely serves as a complement to them. Election observers would be the perfect vehicle to detect unfairness and fraud here. Electoral forensics was not.

Official electoral body National Electoral Commission (NEC)
Election August 4, 2017
Results Official

Leave a Reply