The Republic of South Africa is the most Western — in terms of culture — of the sub-Saharan countries. Best known for apartheid (1948–1981), South Africa had its first free and fair elections in 1994 when it elected Nelson Mandela over F. W. de Klerk and almost filled the National Assembly with African National Congress (ANC) members.
How did South Africa fare in its 2014 elections? Guest writer, Uduak-Obong Ekanem, explores this question.
My first post focused on the new found hope in Africa from the results of the Malawi Presidential elections. Why do I call it new found hope? I will tell you. Africa is notorious with leaders drunk on power and highly resistant to allow democracy win. Rather they prefer to stay in power even if it means doing that through an election that when analyzed will show evidence or consistency with unfairness.
The 2014 Presidential elections in South Africa came down to two parties, African National Congress (ANC) and Democratic Alliance (DA), of which their flag bearers were Jacob Zuma and Helen Zille. Zuma won a total of 62.15% of the votes while Zille was left with 22.23%. While the numbers tell an important story of the winner, numbers also tell a story of the type of election run. Let’s take a journey of the election through the numbers.
Weighted least squares was performed of which the weight was the total votes cast. There are only nine data points to be evaluated — one for each of the provinces. The x-axis corresponds to the candidate support, while the y-axis corresponds to the invalidation rate. A vote is considered valid if it is independent and counts the same regardless of the vote, candidate or party. An invalid vote should be the vice-versa. If there is a relationship between the invalidated votes and the candidate, that might give us some opening to look into the elections further.
The plots above tell an important story. The line of best fit is positive for Zuma and negative for Zille. A horizontal line indicates that there is no relationship detected between the invalidated votes and candidate. The plots above show that for Zille, the lower the votes, the higher the number of invalidated votes and vice versa for Zuma. This is very suspicious if I do say so myself. It seems to imply differential invalidation — in his favor!
A comparance of the p-value of the plot showed consistency with our plot above. One of the things done was to drop the data point for the out of country votes. The reasoning behind this is that, there was little confidence in the out of country votes (OCV). The out of country votes were not necessarily counted (invalidated) in the same manner as the in-country votes. The invalid plot without the out of country data point is below:
There was still no significant change in the relationship of the scatterplots for both candidates that was different from without the out of country (OCV) data points. Conclusions are not final as with WLS analysis, there is need for checking the assumptions to see if the assumptions are violated or consistent.
Assumptions of WLS
The conclusions of the research can only be made after testing for assumptions. The assumptions tested in the WLS include: constant variance, Normality, and constant expected value. Prior to dropping the OCV data, the initial tests of the assumptions all failed the assumption tests in both candidates’s cases.
After dropping the out of country (OCV) data points, there was no violation detected in the assumption tests.
What does that mean? I leave you to conclude. However, with the p-values being so large, I cannot legitimately conclude that there is evidence of unfairness in this election.
This fact is tempered by the small sample size. If I had more data, then I may be able to detect differential invalidation in this election… or not.