# Were the BVR Kits distributed fairly?

The CORD coalition has recently claimed that the BVR distribution has been skewed to favour areas with more JUBILEE supporters at the expense of areas with CORD supporters. the Jubilee coalition and the IEBC have refuted this. This piqued my curiosity so I set out to get my own data answers.

## the Data

The data I used were from multiple sources. These are the sources:

1. The list of election results for all County Assembly ward representatives across the country. Grassroot support for the major coalitions is better represented by voting pattern at the ward level.
2. Shapefiles for the wards and constituency from github. (for Maps)
1. The distribution list of the BVR Kits by IEBC in the ditribution of the BVR Kits.
• I converted all the pdf documents to spreadsheet online here.
1. I had to clean and label parties and their affilieation from information online. For parties for which I could not get information, I tagged as unknown. I used OPenRefine(formerly Google Refine) to merge and clean the data.
2. My final list of parties and their affiliation is as indicated below. P.S (I cannot guarantee accurateness of the tagging, if something is off, please let me know in the comment section)
datatable(data3, options = list(pageLength = 5))

*I tagged each of the parties in the wards dataset to for to tie each of the parties to a coalition. Then I combined the data from IEBC into the shapefiles

aditionaldata = quote( .(total = sum(votes),winning_votes = max(votes), percent =round(max(votes)/sum(votes)*100,0) ) )
sss = sss[,colind := (factor(coalition))]

# Remove some unnecessary data from the shape data and merge with the iebc data

rmlist = c("SPOILT","REJECTED","REPORTED","SPOILT_VAL","VALID","DISPUTED","RESULT","REGISTERED")
places@data[,c(rmlist)] =list(NULL)

# Now combine the data
index <- sss$ward_code #matching key coalition <- sss$colind #get all coalition data from the IEBC dataset
percent <- sss$percent #winner's margin, higher percentage mean very popular MCAs party <- sss$party #winner's party
totvot <- sss$total #winner's party places@data$coalition <- coalition[match(places@data$COUNTY_ASS, index)] #add the coalition column places@data$percent <- percent[match(places@data\$COUNTY_ASS, index)] #add the winner's margin column

## The Analysis

### Background and Assumptions

1. The first map shows coalition strongholds. It is worth noting that the visually, there seems to be almost a fifty-fifty split in the number of wards that are dominated by parties from the noted coalitions.
spplot(places,c("coalition"),main=list(label="Coalition territories")) #showing county Assemblies and their coalition affiliation

1. The second map shows the computed sizes of the wards as computed from the size of the shapefiles. Note that the general area for both CORD and Jubilee are somewhat balanced. Both coalition have some large and small counties.
spplot(places,c("Shape_Area"), main=list(label="Land mass Area",cex=0.75)) #showing sizes of different counties as calculated from shapefile size.

Now, if we are to believe CORD’s assertion that areas that areas linked to it have received less BVR Kits, then we can:

Now, if IEBC has distributed the kits equitably, then we expect:

1. Wards with large area (in sq. KM) and voters/voting stations should have a higher number of BVR kits
2. That (i) above should hold true for both CORD and JUBILEE coalitions

### Comparing CORD and Jubilee areas Results

Visual comparison is sometimes, I think, quite interesting. I will proceed on with visual analysis.

A. The chart below shows the relationship between the computed size of the the shapefiles and the ward size provided by IEBC.

relation <- lm(CAW.AREA.IN.SQ.KM ~ Shape_Area, data= final_data)
plot(CAW.AREA.IN.SQ.KM ~ Shape_Area , main="AREA: IEBC Vs Shapefile", xlab="Computer Generated Area", ylab="IEBC reported Area", data = final_data)
abline(relation, col = "red")

The shapefile projection is longlat meaning that there should be a direct propotionality between the size computed by shapefile and the sizes given by IEBC. From this it appears that some areas, according to IEBC, are larger that they are supposed to be, and some areas are smaller that they are supposed to be.

B. An interaction graph below shows the distribution of the BVR Kits based on the mean sizes of the wards.

with(final_data,{
Issuedkits <- ordered(BVR.KITS)
interaction.plot(Issuedkits, coalition, CAW.AREA.IN.SQ.KM, fixed = TRUE, col = 2:5, leg.bty = "o", fun = mean, ylab="Mean Area in SQ. KM", xlab="No. of BVR Kits issued")
})

The mean is unique in that it has a squashing effect on large values. For instance, if a smaller county and a larger county are given same number of BVR kits, then the average BVR per SQ.KM will be smaller than if same size counties are given same number of BVR. You will note that with smaller numbers of BVR kits are issued to smaller wards in both CORD and Jubilee areas, however, higher number of kits are issued to relatively smaller Jubilee wards.

C. You can further see this with another plot of the same variables, but with median instead of the mean used to compare.

with(final_data,{
Issuedkits <- ordered(BVR.KITS)
interaction.plot(Issuedkits, coalition, CAW.AREA.IN.SQ.KM, fixed = TRUE, col = 2:5, leg.bty = "o", fun = median, ylab="Median Area in SQ. KM", xlab="No. of BVR Kits issued")
})

The median is computed by first aranging values in ascending order, then taking the middle most value. For this chart, what this means is that at least 50% of Jubilee wards issued with 8 BVR kits had an area of about 5000 SQ.KM this is compared to at least 50% of CORD wards that had an area of about 7000 SQ.KM that were issued same number of BVR kits.

D. I also looked at whether the issue of BVRkits was at par with with the number of advertised polling stations.

with(final_data,{
Issuedkits <- ordered(BVR.KITS)
interaction.plot(Issuedkits, coalition, NUMBER.OF.REGISTRATION.CENTRES.IN.THE.WARD, fixed = TRUE, col = 2:5, leg.bty = "o", fun = mean, ylab="Mean No. of Polling Stations", xlab="No. of BVR Kits issued")
})

You will see that while in CORD counties, the number of BVR kits issued was commensurate to the number of polling stations, the same was not true for Jubilee areas where some areas with smaller number of polling stations had 8 BVR kits issued.