Angola goes to the polls

 In Angola, News

There is an undeniable fact. Angolan elections scheduled for the 24 August are the most disputed of the post-civil war (2002)[1]. Never has been a discussion so renamed and the intensity of the arguments and uncertainty so debated.

By CEDESA

Despite some tension and sometimes incendiary rhetoric, this electoral context represents a significant advance in the democratic struggle, which is expected not to go beyond from other forms of struggle.
One of the innovative aspects that has arisen in these elections is the plurality of polls. As much as memory and files allow to determine, the existence of polls was not a usual fact in the previous Angolan elections.
In fact, in 2017, only one reference came up to a supposed poll made by the Brazilian company “Sensus, Pesquisa e Consultoria”. The existence of this poll was never confirmed, but at the time sources revealed that this entity would have learned that the MPLA would gain only 38 percent of the votes. UNITA would obtain 32 percent of voting intentions, while the Casa-CE would appear very close to Unita, with 26 percent. From this it would result that the majority in the National Assembly would be the opposition[2].
What is certain is that this poll has never been confirmed and the final results were quite different. As is well known, the MPLA had 61.05%, while Unita and Casa-CE reached 26.72% and 9.49% of the votes, far from what the supposed poll stated.

2-The polls in the 2022 elections

If the 2017 elections revolved around a ghost survey that had nothing to do with the final reality in 2022, unlocked public polls appear, although fought by the forces that do not like the results.
It is about these polls and the possibility of predicting an end result based on them that this analysis is leaning.

We follow the identification of polls by a Portuguese Cable Channel[3] and consider five published polls. These are:
-Afrobarometer (https://www.afrobarometer.org/)
-Angobarometro (https://www.angobarometro.com/)
-Angopolls (https://angopolls.org/)
-Mudei (https://jikuangola.org/inqueritos/index.html)
-POBBrasil (https://pobbrasil.com/)

All of these entities have their site and present the results publicly.
We do not ignore that there are several controversies around some of these entities, however, we chose to rely on the technical records of each of the polls and the good faith of the interveners. Let’s look at the message and not “kill the messenger.”
In fact, with the exception of Afrobarometer, all other entities are reasonably recent and seem to be dedicated to present Angolan elections, containing professionals from other companies or organisations. This is a sign of democratic liveliness and therefore does not deserve criticism. What is certain is that these entities after the first essay that are these elections in Angola will be perfecting to contribute to the democratic discourse in Angola.
Looking at the data sheets of each of the polls and results we will make a measure of the results following two criteria, the reliability of the method and the normal Gaussian distribution.
First, the type of inquiry performed. From the technical records and affirmations of those responsible we conclude that Mudei and Afrobarometer do random street surveys based on premises that specify in their methodology pages. In turn, Angopolls and PoBBrasil perform telephone inquiries according to a random computed selection. Angobarometro, on the other hand, performs online polls.
We understand that online surveys are not reliable because they benefit from a “neighbourhood effect”, that is, there is a tendency to call friends and people who think the same way to visit the site. Therefore, if a site is taken as closer to UNITA will call more people from UNITA, having a bias in its favour, the same happening if the site is MPLA. To this extent, we believe that online polls demonstrate a party’s ability to mobilise, but not the voting intentions of a random sample of the population.
We thus remove the AngoBarometro from this appreciation.
Regarding the remaining four, we proceed to a Gaussian distribution by eliminating the extremes and maintaining the standard-normal distribution. To this extent we will not consider the POBBrasil Survey that gives the MPLA an extreme victory, and also Mudei that gives UNITA an extreme victory.

3-Standard-polls

There are two polls that seem to us the most standardised: Afrobarometer and Angopolls.
Afrobarometer’s technical record reveals that: “The Afrobarometer team in Angola, led by Ovilongwa – Public Opinion Studies, interviewed 1,200 adult Angolans, between February 9 and March 8, 2022. A sample of this size produces national results with one Error margin of +/- 3 percentage points and a confidence level of 95%. The previous research in Angola was conducted in 2019”.
The results achieved are presented in the table below:

Table No. 1- Afrobarometer results (see image – for all images go to: (https://www.cedesa.pt/2022/08/17/eleicoes-em-angola-sondagens-e-previsoes-de-resultados/)

Angopolls has conducted several inquiries since December 2021. We will focus on the last of the published sequence, “VII-General Elections poll. July 2022”. Its datasheet states that: “The poll was made by telephone. 5040 valid inquiries were obtained, 21.03% of respondents were female.”
The results obtained were as follows:

Table No. 2: Angopolls results (https://www.cedesa.pt/2022/08/17/eleicoes-em-angola-sondagens-e-previsoes-de-resultados/)

¹ Not accounting for undecided and abstentions.
The Angopolls page also contains a curious chart with the presentation of the evolution of results over the several months:

Table No. 3- Voting Trends According to Angopolls (https://www.cedesa.pt/2022/08/17/eleicoes-em-angola-sondagens-e-previsoes-de-resultados/)

In a first analysis, it would seem that Afrobarometer and Angopolls polls give different results. In fact, the presentation of Afrobarometer attributes 29% to the MPLA and 22% to UNITA, while Angopolls refers to a percentage of 60.15% to the MPLA and 39, 85% for Unita. It appears to be a very large difference between the two polls. However, a thinner analysis reveals that this is not the case.
In the end, in essence, the two entities came to very similar results: MPLA wins and UNITA reinforces its result, and even in percentages the difference is not very significant. The explanation for the apparent difference that does not exist is in the methods of presentation of the results and not in the results themselves.
If we notice Angopolls withdraws from their presentation the non-response (abstentionists, undecided on whether they would vote, etc.), while Afrobarometer does not do so. Note that they keep 46% – don’t know, don’t vote, refused to answer.
Now if we apply the same criterion for both polls, that is, removing the do not know, don’t vote, refused to answer, the so -called non-responders we will have a significant approximation between the two polls that mirror the table below:

Table No. 4: Comparative results Afrobarometer and Angopolls Following the same method of presentation (https://www.cedesa.pt/2022/08/17/eleicoes-em-angola-sondagens-e-previsoes-de-resultados/)

It is evident that another method for considering non-response is to impute them according to historical criteria (that is, considering the meaning of voting in previous elections) to political forces or then you can go into several speculative exercises.

Conclusions

What results from this analysis is that normal predictability points to a victory of MPLA in a percentage that oscillates between 54% and 61% and a substantial reinforcement of UNITA to 40%, with a sharp decrease of other political forces, what is called in political science a bipolarisation.
It should be noted, however, that giving the percentage of non-respondents, these numbers are not fixed and definitive. They are a photograph at a given moment, but everything can change.
Therefore, polls give some indications, mark trends, but do not give certainty. As mentioned at the beginning of this work, this is an area of democracy that only now begins to be explored, so you can’t expect definitive answers, but only mutable observations.

[1] In this analysis it is not included the 1992 elections because they were conducted in a historically very different context.
[2] https://www.makaangola.org/2017/08/sondagem-eleitoral-mpla-fica-atras-da-oposicao/
[3] CNN.pt
[4] Due to rounding adjustments it doesn’t equal 100 %