massive mahinda victory to close election? how sunday leader columnist indi.padashow twisted election projection data

on january 7th indi.ca blog (run by the sunday leader columnist who also runs the scurrilous padashow blog) published a "hypothetical analysis ...on what could happen" at presidential election. the conclusion of that election projection was in his own words "very hypothetical, but looks competitive" with both major candidates getting close to 50% of vote.

in this post we will look at the data in this "analysis" and show that a 'competitive' / 'close' (as he keeps on repeating his posts) election can only be projected by manipulating data to suit that outcome.

without such obvious manipulation even his data shows a massive victory for mahinda rajapaksa.

i am making available a screenshot of the post and a copy of word document attached to that post separately in case anybody decides to change things.

i also made a comment there (which said most of what i write here, in brief) and was censored as usual. his censorship is personal. he does allow some critical comments by others but nothing from me. since i don't make comments under any other name all my comments get censored. of course he has a right to do as he please in his blog.

before going into data we have to ask, who did the 'analysis'?

not indi.padashow. he has trouble calculating simple percentages and having grown up in foreign countries is generally ignorant about most things sri lankan. this is one of his recent statements about the election. (click to enlarge)

lol. anyone who has even heard about the law or has seen the ballot and voted in a presidential election here, knows that it is possible to indicate 3 preferences.

funny isn't it, that such an ignoramus is presuming to lecture sri lankans about law and politics. funnier still that so called 'newspaper' sunday leader ( which has taken a management decision to back sarath foneka ) would employ him as columnist. even funnier that unp used this person, son of a political appointee (corrupt by son's standard for accusing others of corruption), to run their online campaign (nothing more than a website really ) in 2005. (he used to actually edit comments posted in that unp website so that even unfavorable comments appear favorable. not my comments, but then frequent blogger morquendi's, hardly a friend of mine). [indi does not deny any of this if pressed (given the evidence in his own blog and others, he can't anyway), but is not into voluntary full disclosure either as demanded by journalistic ethics. iow a worthy and fitting columnist for the rag sunday leader.]

if not indi who made the election projection?
in the post he says "i think the document is from parliament somehow". my request for more information on that point went unanswered.

the properties of word doc attached to post says it was authored by "ajiwadeen_m". we have no idea who that is .

there is also the fact that same numbers and same documents appeared in at least two or three other places (including some pro ltte terrorist blogs/websites, not the ones blocked in sl). indi makes no reference to this in his post though at least one was clearly earlier than his.

indi also tweeted this,


since he does not provide the details of author, but published them on his blog, we will consider them his, until we get further information .

a look at the data.
i will confine myself to most obvious unexplained discrepancies.

he provides two forecasts based on ethnic and district votes (forecast I & II). lets start with first one.

Major Ethnic Groups	Valid	MR		SF
	Votes	Votes	%	Votes	%
Sinhalese	8,424,360	4,633,398	55.0%	3,731,991	44.3%
SL Tamils	868,198	182,322	21.0%	607,738	70.0%
Muslims	1,039,375	291,025	28.0%	744,193	71.6%
Ind. Tamils	618,859	326,138	52.7%	290,864	47.0%
Others	33,090	12,905	39.0%	19,854	60.0%
TOTAL	10,983,881	5,445,788	49.58%	5,394,640	49.11%

there is no explanation as to where he got these numbers.

according to this muslims comprise 9.4% of valid votes. according to department of statistics and they make up about 7-8% of population.

imo the figure for sl tamils is way too high and indian tamil figure way too low. we do not know true figures bc 2001 census is incomplete due to terrorist control of several districts then. as such there is no valid source for this sort of a breakdown.

for now we will proceed with these rather shaky looking figures

muslim vote
he says 71.6% of muslims will vote for sarath fonseka? why? he does not give a valid explanation even in word doc. muslim politicians and parties are split. if anything more elected muslim mps support mahinda rajapaksa than oppose him. nor is there any proof of such an anti mahinda tendency among muslims in recent elections.
a more realistic and rational projection would split the vote 50/50, or 45/55. nobody rational will go beyond 40/60. but indi padashow projection does.

now assume it was 40/60, instead of 28/72, among muslims, what will be the results? 1-1.5% increase in mahinda total vote percentage with similar decline in fonseka's

indian tamil vote
( i prefer to call them up country sri lankan tamil but will stick with this name for now)
here he says vote will split 52.7/47.
now ceylon worker's congress (cwc) the main political party of indian tamils, as well as second biggest, upcountry people's front (upf) are supporting president. only few defectors from them are with fonseka. and historically cwc has delivered 70+% of the indian tamil vote. in every election it had defections (usually the same ppl) but that does not seem to hamper its vote delivery. but indi padshow's number cruncher assumes for unknown reasons this time will be different. why? no reason given

now assume this vote splits conservatively at 60/40, instead of 53/47 as in indi projection, or more realistic 70/30, what will be the results? 0.5+% increase in mahinda total vote percentage with similar decline in foneska's.

these two almost certain but still very conservative roll back of unexplained irrational manipulations will result in a 51.5/47 split in total votes. a 4.5% or about 500,000 vote margin. realistic vote splits among muslims and indian tamils will double even that .

sri lankan tamil vote
i am not going to say anything about sri lankan tamil vote split. i don't know how it will split. it could be same as this prediction says(21/70/9) or it could be 40/50/10 or even 45/45/10.

anything is possible, so i am not going to speculate on any manipulations there.

ordinary tamils here, in india, or in diaspora, have generally rejected racist terrorist supporters ( like tna/itak) by not voting or voting for others if they get a chance .

as i said before i don't think sri lankan tamil vote will be high as these numbers. quite possibly it will be lower than indian tamil votes.

but lets leave him with his 21/70/9 split without change .

sinhalese vote
here he gives a 55/44 split. in recent provincial council elections the split in mostly sinhalese areas was 68-70/31-29.

however this is best examined by looking at his district based forecast (forecast II).

let us concentrate on kurunegala district (with no special connection to either candidate)

this is the last provincial council election result -
valid votes 709,186, upfa 497,366 (70%), unp+jvp 209,632 (30%)
in the padashow projection -
valid votes 938,397, mr 495,474 (53%), sf 436,355 (47%)

in other words, valid votes go up by 230k (which is ok since more ppl vote in national elections than in provincial elections),
but upfa vote fall by 2k (and 17% as share). almost all the extra votes go to sf, increasing his vote share by 17%

is that realistic? of course not.

reasons for this manipulation?
in word doc he says he reduced mahinda's vote in each district by 10-15% from what upfa got in 2008/09 provincial elections. no explanation why a major party will get less votes from a larger number of valid votes in a national election than in a provincial election.
he also says that since unp doubled its vote numbers from 2004 provincial election to 2005 presidential election, he is assuming it will do so again.

here are the actual figures for kurunegala
2004 pc election - valid 604,979, upfa 362,084, unp+slmc 240,078.
2005 presidential election - valid 896,497, mr 468,507, rw 418,809.
valid votes went up by 290k, upfa votes went up by 105, unp went up by 178k.

iow, he replicates 2004-2005 unp vote increase in pc to presidential elections, but does not want to do likewise with upfa's. bias much?

now let us assume that 2004-2005 scenario gets reenacted without the irrational manipulation .

we assume that valid vote increase from 2009-2010 is as he says, 230k, then instead of all of that going to sf as in indi projection, only 2/3rds goes to sf as in 2005. that would be about 154k. upfa will get other 1/3rd, 76k.

then the results for kurunegala in 2010 will be, valid votes 938,397, mr 573,000 (61%), sf 364,000 (39%) (in every step i rounded up for sf while rounding down for mr).

this is a percentage difference of 8% and in number of votes, 77k (compared to indi projection) for kurunegala .

same with other districts. it is less observable in districts where pc elections were held in 2008 before the war ended, but still there.

so even if we add only 1/3rd of increase in vote from pc to presidential elections to mr, and add the rest to sf , we get an 5%+ increase in mr vote share, over what indi.padashow gave mr in his projection.

this conservative change applied to his all island district based forecast (forecast II) results will give us a overall vote split of 53/46.
remember this accepts without question his figures for north and east provinces, as well as colombo and central (which as you may guess were similarly manipulated but will require lot of space to untangle, so i am not doing that here)

if we apply the above logic to sinhalese votes in forecast I, we get a 60/39 split in that segment. adding 418k (3-4%) to mr's total vote share, while reducing sf total by that amount.
even if we, to be on the safe side, reduce it by half, we still get 57/42. adding 210k (2%) to mr vote share while reducing sf's by same.

if we combine sinhalese, muslim, and indian tamil, sanitized vote numbers we get a overall split of 53.5/45. a 900k, 8 .5%, margin .which can hardly be called 'close' .


here are the sanitized results for forecast I - 53.5/45
here are the sanitized results for forecast II - 53/46

imo the vote margin will be bigger, as i have been saying from the start

remember i have not sanitized most of the manipulations, only the most obvious and easily demonstrated ones (esp in forecast II). i have also not sanitized his initial questionable valid vote numbers.

also as anyone who followed this carefully would have realized, when sanitizing manipulations i took a very conservative route (from rounding down votes for mr & up for sf, to cutting the possible mr increases by half )

given all that, it is safe to say even numbers of sarath foneska's devoted supporter indi.padashow, stripped of obvious manipulations, indicate that sf is in for a trashing.

෴

ps
by all means point to any errors on my part if any .
pss
i wonder if i get banned from kottu this time, for exposing indi padashow 's doings again. even though there is nothing false here. lol

my twitter - http://twitter.com/sittingnut
this blog's twitter - http://twitter.com/llibertarian