Center for Electoral Forensics

The Uganda Show

[Ugandan Flag]The most-recent Ugandan presidential election was held on February 18, 2016. This marks the second Ugandan election we have examined (see: first and first revisited). As we recall, the 2011 election had several clouds over it, including the fact that the government quickly took down the election results.

Is the latest election a paragon of free-and-fair? Guest author Uduak-Obong Ekanem explores this election.

[Yoweri Museveni]

Yoweri Museveni, the current President of Uganda

Already, I have analyzed two elections: Malawi and South Africa. From these two, we have learnt a lot. The important question still remains. Is there really fairness or credibility in elections conducted in Africa? Can the elections reflect the true voice of the people? Or is it the voice of the manipulator, the greedy one whose sole intent is to have power, riches and control. These questions and many more all want answer. To answer this, it is important to ask, what does the Ugandan show say?

The 2016 Presidential elections of Uganda were contested among eight candidates. The elections came down to two individuals: incumbent Yoweri Kaguta Museveni and rival Kizza Besigye Kifefe. Museveni came out victorious winning 60.62% of the votes, with Kifefe winning 35.61%. Now, the public might have accepted the results calmly but as statisticians, trained to look beyond the mere numbers, we will dig deeper into the elections and let the numbers paint the true picture.

Invalidation Plot

Weighted Least Squares (WLS) examines the relationship between the transformed invalidation rate and the candidate support, while weighting with the total votes cast in the district. Prior to plotting, there was a district with 0 invalidated votes. This was, prima facie not believable. Regardless, the specific data point was drop in the analysis. The scatterplot of the transformed invalidated votes in the y-axis and candidate support in the x-axis was done as shown below:

The plot shows something very important. The line of best fit is an indicator of how the variables correlate. The plot above have horizontal lines. The plot are thus interpreted to mean that there is no detected relationship between the two variables of invalidated votes and candidate support.  The line of best fit suggests to us one thing: there is no detected relationship between the two variables.

WLS Assumptions

However, statistical analyses are based on assumptions, the assumptions of constant variance, Normality, and constant expected values were be tested. With the weighted least squares model, none of the tests indicated assumption violation. Thus, this weighted least squares model is appropriate for this data.

The conclusion, based on this analysis, is that I was not able to detect evidence of this type of unfairness in this election.



Leave a Reply