Voters' literacy levels and the number of rejected votes in their constituencies

You know I like playing with data. Actually, that’s what my work entails. I have to make sense of some array of data and either validate a previously hypothesized theory or reject the same. Occasionally, the data may not produce any meaningful information. This could be as a result of many things. It could, for example, be caused by human error - the most common cause. It’s courtesy of this that we have the infamous GIGO - garbage in garbage out. If the data clerk, or even the analyst, makes a mistake, then that would in most likelihood result in a bias. This would thus, distort the conclusion inferred from the analysis.

In some instances, there could have been mistakes at the data collection stage; either due to a poorly produced tool (questionnaire etc) or the method of administration. Rarely, does this result from a faulty computer or software. When the latter occurs it could be as a result of a bug in the software or even a computer virus.

Deliberate distortion of data can also lead to a bias as well. This may be influenced by some ulterior motives, where the user selectively chooses what to include in the analysis and what not to include.

Anyhow, while reading the online version of the Daily Nation, I came across some breaking news saying that the Kenyan referendum results have been gazetted. This somewhat sprung my statistical curiosity into action.

Since I had preciously visited the IIEC website, I thought I should just check the results once again. Just in case I ‘discovered’ something useful. Anyway, I was happy to see that the results are easily accessible in a .pdf format. I therefore, saved the results on my desktop and exported the tables to an excel sheet.

The first thing that I noticed once I moved the data to an excel sheet was the ‘rejected votes’ column. I wasn’t looking for anything in particular but just happened on that. Anyway, it didn’t take me long to see some negative correlation between the assumed illiteracy and poverty levels of a constituency and the low number of rejected votes. Before you run away, let me explain what I mean by this technical term.

In negative correlation, as the value of one variable increases another somewhat related variable decreases. By variables, here, I mean the poverty and illiteracy levels on the one hand and the number of rejected votes on the other. However, this may not be true if you check for a relationship using the percentage figures. You can only see the association if you use the absolute number of rejected votes.

Please note that I do not, as of now, have data to validate my assumption about the levels of poverty and illiteracy among the constituencies that recorded the least number of rejected votes. I based my hypothesis on a recent study that I read on the Daily Nation portraying NEP as the province that, unfortunately, leads in these two areas. The same was true of the districts in ‘upper eastern’ and some other pastoralist’s areas. I remember Nyanza was also considered as one of the impoverished regions in the country.

Thus, when I saw constituencies mostly from the above regions having the least number of rejected votes, then I thought the reason for this could be the poverty and illiteracy associated with these areas. Here, I am talking of all the constituencies in NEP with the exception of Dujis having less than 100 spoilt ballots. The expansive arid north was also well represented with Turkana North, Moyale, Samburu East, Saku, Laisamis, Isiolo South and North Horr all having less than 100 rejected votes. However, for some reasons, Kathiani and Vihiga are also represented in this category of constituencies with the least number of rejected votes.

Anyway, I can only hypothesize that the higher the literacy level in an area, the more the illiterate voters would want to fit in. They might probably feel ashamed in asking for help during the actual casting of the ballot and thus making them prone to commit some mistakes. On the other hand, the voters from the regions known for their low literacy levels would have no problem informing the IIEC personnel of their inability to vote unaided and thus reducing the probability of rejected vote. This could be the reason behind the low number of rejected votes in the least literate constituencies. Conversely, Starehe, Langata and Embakasi having the highest absolute number of rejected votes.

However, this might not be the only valid conclusion. One other reason could be the high number of voters registered in a constituency and could certainly increase the probability of rejected votes. I will test this last hypothesis in my next article.

Refuting Hon. William Ruto's statistical fallacies

Following the just concluded referendum, it is evident that Kenyans have overwhelmingly accepted the Proposed Constitution. The proponents of the ‘No’ side, on the other hand, have also ‘conceded defeat’. However, their concession has come forth with some statistical fallacies that should not be left unchallenged.

Hon. Ruto was quoted in most media outlets, in the country, inferring that close to 60% of Kenyans either out rightly rejected the Proposed Constitution or abstained from voting. However, one key issue that all Kenyans need to note is that the referendum, just like all the preceding opinion polls, showed that a significant majority of the people of Kenya are in favour of the new Constitution. Of all the polls that were conducted, the most valid and accurate should ideally be the referendum itself, which has a much larger ‘sample’, read voters.

Earlier, it was reported in the media that the registered voters for the referendum were over 12.5 million. And the results showed that 70% of those who participated in the plebiscite voted for the new Constitution with 30% voting ‘no’. However, the 70% and 30% respectively represents the universe of those who voted only. Those who abstained from the referendum were about 30%, according to the IIEC. A distinction, therefore, needs to be made as regards these two similar but separate 30 percents.

In combining both the proportion that abstained and the ones that voted ‘red’, the ideal thing is to recalculate using the absolute numbers, rather than incorrectly adding the two percentage figures. If we add the percentages, the ‘60% that either abstained or rejected’ the Proposed Constitution’ and the 70% that voted ‘green’, then you end up with an accumulated 130%. You do not need to be a statistician to know that this is impossible. If we recalculate using the total registered voters, we find that 48% voted ‘yes’ and 22% voted ‘no’ and a 30% that abstained. These correctly add up to 100%.

Since we know that the referendum was the most valid and accurate poll; then, using the same 70 and 30 percents, we can correctly come to the conclusion that, of those who abstained, 21% would have voted ‘yes’ and only 9% voting ‘no’. These last two figures are respectively, 70% and 30% of the 30% that abstained. Therefore, using these recalculated percentages, we end up with about 69% ‘green’ and 31% ‘red’. In absolute figures, these would mean; 6 million who actually voted for ‘yes’ and another 2.6 million hypothetically voting ‘green’. On the flip side, it means 2.8 million Kenyans who actually voted ‘no’ and another 1.1 million hypothetically voting ‘red’. The total, both for the absolute figures and the percentages, add up to the number of registered voters (12.5 million) and 100% respectively.

I need to also caution the propagators of these misconceptions from attempting to use the 48% figure written above as below the ‘internationally recommended 50% threshold’. This figure only comes up when you try to cater for the abstained votes, among the universe of the registered voters. Therefore, the correct thing would be to use the same (48%) while acknowledging the proven breakdown of ‘yes’ and ‘no’ voters among the abstainers – which, as stated, is 70% and 30 % respectively. Thus, you either accept the 48% with the additional hypothetical votes or else leave it as it is currently – 70% actually voting ‘yes’.

Also, we can, with statistical certainty, predict that the so named hypothetical voters above, given a chance, would have voted exactly in the manner as envisaged above. This is informed by the fact that, as written above, the validity and accuracy of the plebiscite is above reproach. Therefore, if we had other Kenyans who being eligible to vote but did not register; then they too would most likely have voted 70% for ‘yes’ and 30% ‘no’.

Using the same distorted logic circulated - that of 60% either rejecting or abstaining from the referendum - we can as well come to the conclusion that 70% of the registered voters turned out for the referendum and 70% of them voted ‘yes’ adding up to a total of 140%. However, it would be inappropriate for me to refute one statistical fallacy with yet another one.

Therefore, the only valid conclusion is that the majority of Kenyans are in favour of the Proposed Constitution, regardless of whether they registered for the referendum, voted on August 4th, or whichever side they voted for or would have voted for. In other words, if your get a fairly representative number of Kenyans anywhere, they would have voted 70% for ‘green’ and 30% for ‘red’. We have no reason, at this point in time, to believe that those who abstained were a homogenous group; from the same region, with similar ethnic and religious background etc.