And the Winner is…. All Under Heaven

Once upon a time…

With 47.4% of the vote (plus or minus 2 percent) the incumbent, Mahinda Rajapaksa relinquished the presidency to the challenger Maithripala Sirisena.

Really?

Well not quite. This is based on assumptions that may not be borne out. Using a simple model of the electorate and simplifying assumptions I arrive at various conjectures. All seem possible not only a MR win but also a MR defeat is possible as well. Useless model? No not really.

When the faith based vote proportions are kept constant and the proportion of loss of Buddhist votes is made zero, (same proportion of the vote won as in 2010 applied to the estimated Buddhist voter base) then the MR vote proportion reaches above 50% of the expected vote. MR wins despite all the cawing, dancing and jumping around.

Is this real? Well it can only be real after the votes are counted and announced on the 9^th of January 2014. Till then this is conjecture.

That’s how the numbers stack up right now when I use a model with what I consider to be plausible assumptions. What is plausible to me may not be for you. This has nothing to do with what I want or wish for. Using reasoning that will be proved to be absurd, flawed or prescient or simply all right, in a few days my work will be, toast?

Model 1 – Faith based voting[1]

Assumptions

1. Proportion of Vote by faith for MR.

Hindus, 20%

Islamic 15%

Catholic 40%

Other 40%

2. Proportion of Buddhist vote in all electorates is 5% lower than the total winning proportion for that district in 2010.
3. The voter turnout by electorate is the same as 2010 with an adjustment for the Northern province keeping in mind that the voter base was overestimated in 2010. The voter turnout is set at 70% for the north.
4. The voter population proportion of the district by faith is the same as that of the human population.
5. The voter base for each district is that of the new electoral list set for the presidential election.
6. The rejected votes will be 0.97% as in 2010.

The proportion of the vote that the incumbent receives based on these assumption is – 47.4% not enough to secure victory.

Change the assumptions and presto! The results change a little.

How sensitive is this simple model to the change in the assumptions in this simple model?

If MR maintains the proportion of the Buddhist vote in each district at the same level as the total vote proportion he received in that district in 2010, he will win over 50% of the valid votes to be cast, holding all other assumptions constant. ( In predominantly Buddhist Sinhala electorates that is where the Buddhist population proportion is higher than the national average of 70% the proportion of the total vote received is obviously predominantly Buddhist, I think this is a safe assumption).

Obviously if the faith based vote proportion increases over the assumptions made here then the decline in the Buddhist vote can be above zero and still give over 50% of the votes cast to the incumbent.  The faith based vote has to increase by 5 percentage points and the loss of Buddhist votes has to be 2 percentage points below the district winning vote in 2010 in order for the MR total to be over 50%.

The outcome remains unchanged at 47.4% of the valid vote for MR when the voter turn out is increased or decreased by x%, holding other assumptions constant. However the result changes if the turnout is changed in some specific areas and not in others.

In particular, if the turnout in the North and East is lower than the assumed 70% and the proportion of the vote to MR in the NCP is lower than the proportion based on the vote of 2010. The proportion of the vote to MR falls by 0.35% to 47.05%

Will I change my mind? Will you if I am wrong?

Model 2. New vote and old vote[2]

Another model and different results. I must be kidding? No!

In this model if the new voters; the youth, first time voters, are for MR (the Namal effect) more vigorously than 40% of the total and if the loss of votes from his vote “base” in 2010 (unmet expectations) is less than 10% hello President MR, again!

There are assumptions of course, always assumptions.

Assumptions

1. The net addition to the voter base is assumed to be the Namal effect, the youth and the young voter. MR proportion of this vote is assumed to be 45%. The actual new vote is expected to be larger than the net addition; the loss of the voter base in 2010 due to death is factored in.

The new vote is expected to be larger than the net change in the voter base between 2010 and 2015. First part of the voter base of 2010 would have died.

2. Using the life tables for Sri Lanka I estimate the death rate to be 5 per 1,000 for the 2010 vote base. However, the voter base in 2010 is an overestimate due to the number of voters on the lists in 2010 from the northern province. A comparison of the voter base for the Northern province shows that the 2014 base is significantly lower than in 2010. This is a difference of 20% of the 2014 voter base or over 200,000 voters less in 2014 than in 2010.

3. The vote that MR received in 2010 for each district is reduced by a fixed percentage of 5% in all districts. In the NCP it is reduced by 10%.

4. The voter turnout is fixed for each district at the same level as in 2010 but increased to 70% for the Northern province.[3]

MR vote proportion remains above 50% of the valid votes for parameter values with the Namal effect not lower than 40% and loss of old vote not higher than 10%. This is a robust result over a range of combined values for the Namal Effect and voter loss, the unmet expectations effect.

You want me to show my models? What will you show me? Don’t please! Please Don’t.

So will there be a change? It’s possible.

“All under heaven! ” This is something to think about, it needs some more time .[4]

[1] The faith based model uses the electoral base for 2014 and the voter turn out at district level. I apply a proportion of the faith based population for each district from the last Census from the Department of Census and Statistics

http://www.statistics.gov.lk/PopHouSat/CPH2012Visualization/htdocs/index.php?usecase=indicator&action=Map&indId=10

to arrive at an estimate of the faith based voter base. I use the parameters of voter turn out and the faith based MR vote assumed as given and calculate district level faith vote. In the case of the Buddhist vote after calculating the base I apply a slightly different assumption than for non Buddhist faith which have a fixed MR vote proportion. The Buddhist vote changes by district.

[2] This model is different from the first one. I derive the new vote in each district by deducting the 2010 voter base from the 2014 base except in the case of the Northern province where the 2010 base was over estimated. I assume that the new vote will turn up like the old vote and use the 2010 turnout for all districts except those of the Northern Province. There I use a proportion of 70%. I also add to the difference between the new and old voters base the estimated loss from the 2010 base due to death. The Namal factor is then applied to the new vote, i.e. the proportion of the new vote to MR. I then take the vote that MR received in 2010 by district, reduce it by the estimated deaths, and then apply a loss factor. The two numbers are added, totaled by district. This is the total expected MR vote. The ratio of the vote to the total valid vote expected is calculated. The assumptions are different and so are the results. What matters is the strength of the Namal effect and the loss of votes from the 2010 vote, the unmet expectation effect. This is also present in the first model.

[3] Here I make an assumption of excluding the Northern province from the base to calculate the addition to the voter base. This is because the population and the voter base of the northern province in 2010 are not well estimated. We have numbers but they seem to be subject to much error. The voter base of the north in 2010 was not an updated list and there seem to have been many losses due to war, displacement and people leaving the north. This is a saga that I will not go into here. A simple basis for computing the new vote is to look at the age cohort that will pass the threshold of the eligible to vote age of 18 since 2010. This is a cohort of size of approximately 1 million, but the details are important.

[4] For those who have not seen it I recommend the movie “Hero” directed by Zhang Yimou. And the statement, “All Under Heaven”? It is not easy to interpret.