|
Post by reporter911 on Jan 11, 2013 21:40:35 GMT 3
The glaring methodological weaknesses in the poll aside, I just looked at the date of the polling and noticed that it was done even before Musalia put together the Amani Coalition and Eugine stood down for him in the presidential race. Go advise your rumored statehouse protégé Mudavadi to pack up and go look for votes in his Nyeri backyard!! Western Province iko na wenyewe Kapiche!! Mudavadi ( remotely controlled statehouse Protégé like Uhuru Indicated stands no chance in Western province.. 99% of luyhia's already know he is a puppet ;D ;D[/size]
|
|
|
Post by reporter911 on Jan 11, 2013 21:41:46 GMT 3
Mwalimu, Why should I give my own scientific polling yet I'm not an expert in that field? I challenged those who are discrediting Inftotrak to go out to the field an carry out their own research. Unless someone can provide an alternative scientific polling with different results, I have all the reason to rely on what's been published and treat it for what it is! Rais,I do write in very simple English which should also be very easy to understand. I did not ask you to do any scientific polling, but to give us your scientific analysis of the figures Ambitho has already provided. I am glad Abdulmote, being the sober guy he always is, has already relooked at this thing and come to the conclusion that it is a very weak and quite frankly, a very unreliable poll. Very Reliable except for those supporting Mudavadi/Jubuilee ;D ;D na mambo bado
|
|
|
Post by raiswakesho on Jan 11, 2013 22:00:19 GMT 3
Well, well Bwana Mwalimu, I don't know how one can give a scientific analysis on the polling without recreating the models & everything that goes into this kind of work. I also lack the expertise to even audit the same, if I may add. You may disagree with the polling because it doesn't favor your candidate but to dismiss it as fake without proper substantiation is by itself silly!
|
|
|
Post by bkichwa on Jan 11, 2013 22:05:59 GMT 3
I noticed there is the usual dispute on how scientific these polls are. I appeal to those who are meticulous on methodology and variable definition thus: Help me out folks: Over the last six months I have reached the conclusion that, whereas they are other variables influencing the choice of PORK, ETHNICITY will be { outside of Nairobi} the determinant factor amongst the big ethnic groups the GEMA, the Luo, the Luhya, and The Kalenjin. I therefore miss an opinion poll which breaks it down for me: 1. What percentage of Kikuyu voters will vote Jubilee? 2. What percentage of Jaluos will vote Cord? and so too Kalenjin and Luhya. I did not study mathematics, but I think ETHNICITY may be, in this case in Kenya of the March 4 election, a better scientific indicator of the voting preference, than the other ' statistically representative Kenyan'. I doubt there is anything definable as Kenyan in you mates! that is worth a scientific variable: that would be a holy grail tale! The representative Kenya can then be confined to the minority tribes who give no damn anyway, coz they have no hope in hell of getting their man to the top job. Consequentially with a low turn out. Now, has any Jukwaaist given a dense amount of thought to using ethnicity as a scientific yard to predict party allegiance? [That is what zoning is all about anyway, but do we say!] Abdulmote, In response, see Jakaswanga's post above. I find it unwise to ignore the standard voting behavior of Kenyans since the advent of multipartyism in 1992. Use the latter to analyze who might get votes where and you can get a sense of how this thing is likely to play out. Yes, some of us who form Kenya's middle class might intellectualize and introduce other theoretic factors that shall be possible determinants of who votes how, but the reality remains that majority of Kenyans still.are guided to voting by the very same reasons that have existed since 1992.
|
|
|
Post by mwalimumkuu on Jan 11, 2013 22:20:12 GMT 3
Well, well Bwana Mwalimu, I don't know how one can give a scientific analysis on the polling without recreating the models & everything that goes into this kind of work. I also lack the expertise to even audit the same, if I may add. You may disagree with the polling because it doesn't favor your candidate but to dismiss it as fake without proper substantiation is by itself silly! Since in your own words you lack the capacity to understand how these things work, it will be a waste of time to continue engaging you. Keep to those generic responses and lets move on.
|
|
|
Post by raiswakesho on Jan 11, 2013 22:37:51 GMT 3
I thoughtt I was the one who couldn't understand plain engilsh but anyway, you are entitled to your own opinion but not the facts. Show me the facts and I'll offer my comments. By the way, your are not forced to respond to my posts if you think it's a waste of time...Do I have to remind you on that?
|
|
|
Post by mwalimumkuu on Jan 11, 2013 22:47:29 GMT 3
I thoughtt I was the one who couldn't understand plain engilsh but anyway, you are entitled to your own opinion but not the facts. Show me the facts and I'll offer my comments. By the way, your are not forced to respond to my posts if you think it's a waste of time...Do I have to remind you on that? You didn't have to because we engage here with a view of progressing in our daily learning. But when one realizes that they are just marking time with inanities without making any meaningful progress, what other option is left if not to disengage?
|
|
|
Post by mank on Jan 11, 2013 22:52:01 GMT 3
I noticed there is the usual dispute on how scientific these polls are.
I appeal to those who are meticulous on methodology and variable definition thus: Help me out folks: Over the last six months I have reached the conclusion that, whereas they are other variables influencing the choice of PORK, ETHNICITY will be {outside of Nairobi} the determinant factor amongst the big ethnic groups the GEMA, the Luo, the Luhya, and The Kalenjin.
I therefore miss an opinion poll which breaks it down for me: 1. What percentage of Kikuyu voters will vote Jubilee? 2. What percentage of Jaluos will vote Cord? and so too Kalenjin and Luhya.
I did not study mathematics, but I think ETHNICITY may be, in this case in Kenya of the March 4 election, a better scientific indicator of the voting preference, than the other 'statistically representative Kenyan'.
I doubt there is anything definable as Kenyan in you mates! that is worth a scientific variable: that would be a holy grail tale!
The representative Kenya can then be confined to the minority tribes who give no damn anyway, coz they have no hope in hell of getting their man to the top job. Consequentially with a low turn out.
Now, has any Jukwaaist given a dense amount of thought to using ethnicity as a scientific yard to predict party allegiance? [That is what zoning is all about anyway, but do we say!] Even in the study design you are proposing we would still want to see a more encompassing casting of the sampliing net. And you are right, ethinicity is the most important defining factor of the typical Kenyan's voting inclinition. The simple static design mostly used by statisticians (like the topical one) is premised on the idea that simple stastical measures (mean, standard deviation and the like) of outcomes in a representative sample are transferrable to the population (key qualifier: representative sample). This premise is that if your sample reflects the population, then you do not need to isolate the individual effects of the variables defining population behaviour .... you study the sample as if it is a full cross section of the population. so it is obvious what happens when you decide that a certain ethinicity (e.g. Embu) need not be sampled (for Eastern Province). The method you are proposing is more ambitious: it seeks to isolate the individual effects of various explanatory variables for the outcome of interest, and then to apply those effects to the defining variables of the population to come up with population effect. That's the mainstay of economists (econometrics). Either way, the representativeness of a sample is important, in that if you miss to sample enough study units with an important determining factor, your results will be suggesting that such a factor is unimportant. ... but the second method, is less dependent on exact representativeness of the population ( i.e. it does not asume that the sample is like full crossection of the popluation, rather it assumes that all the important decision variables of the population are sufficiently represented in the sample) - it requires, of course, that the sample elements on average look like those of the population (i.e. you are not sampling only the extremes - and it turns out that random large samples will generally furnish this quality), and that there be enough study units with expression in each of the variables of importance (you might have had of some jargon like degrees of freedom, system identification, etc - all those come into play here) so that individual variable effects can be mathematically extracted. In this case the authors choose to use the static method which requires sample likeness with population, and then they do not include all ethinicities (yet we know the importance of every ethnicity). When studies do this, often they make the assumption that some ethnicities are like copy cats of others, e.g. they might say that once you have sampled the Meru you do not need to sample the Embu. That might work in certain cases ... not so elections. In fact even Meru by itself is very polarized, and if sampling in Meru does not balance that polarity it will end up biasing the result, suggesting that Meru is more like the locality it studies more ... and in this case, it will extend the same biase to Embu and to Eastern province in general.
|
|
|
Post by raiswakesho on Jan 11, 2013 23:07:08 GMT 3
Mwalimu,
I'll let wanajukwaa be the jury and make their own assessement as to who has shown a lot of emptiness in this forum! I don't play in that league bro!
|
|
|
Post by Sam Ochola on Jan 11, 2013 23:43:52 GMT 3
Infotrak poll concurs with the in-house survey I just carried two days ago, although my findings put Uhuru closer at 45% rather than 39%. There had been a surge of hoi polloi numbers chant all over the place referenced to the IEBC voter register. I decided to get behind the scenes, "behind the enemy line" to analyse the ground. First, the futility of MM's candidature has clearly been exposed and consequently his supposed Luhyia base is overwhelmingly gravitating towards CORD. However he still have remnant support, 25-30% home votes, of which if clings to might curtail RAO's outright win in round 1. Nonetheless in the event of a run-off, all these "displaced" western votes will mostly likey align to RAO, with or without MM's blessing. It goes to say that subject to how he handles the already Messy situation (like his VP pick) Mudavadi's bid could metamorphosize into a total joke where by come the ballot day his candidature will be "pro non scripto" ie negligible even among his Luhyia diehards. That is when he will finish with a Shikuku/Kukubo votes as opposed to Kijana Wamalwa's (1997) that he's currently envying for the sake of a bargain.. In the aforesaid scenario, CORD becomes a unanimous alternative of western votes; a decisive support that would most certainly put Raila home & dry in 90 minutes. Second notable issue about this election is the Kalenjin's Rift Valley votes ( registered at 1.5M). The jubilee (read TNA ) numbers' bravado is HEAVILY pegged on the assumption that this vote will be herded in-toto to face mt Kenya; because Ruto is VP and the "promise" of 50% govt share...Oh yes a promise, a binding MoU! This is where the biggest surprise awaits the dark forces. As a fact on the ground, the ODM/ Arap Mibei wave that swept the Kalenjin political landscape in 2007 will neither be swapped nor be replicated this election cycle. Equally, although no one is discounting the existence of Uhuru's support among the Kalenjin, there are some known unknowns here! If I'm in jubilee, I wouldn't toss a champaigne yet over Rift valley votes beyond Nakuru & Laikipia. Achicha Wendi Kityo is a brewing undercurrent that could spoil the mountain party. My statistical observation is that 30% Kalenjin vote evading Mr Kenyatta will leave him in deep hole. Apportion that number to Raila and the party could be over. This is only the Kalenjin constituency without factoring the Maasai, Samburus & the Turkanas of Rift-Valley. In other words a modest support from Ruto's stronghold is just enough for CORD riding forward given the overall picture of the race. Mambo bado! ABR scheme through MM & western has seriously boomeranged. I reckon dark forces are nocturnal in nature and they never sleep. But evidently there are running out of time and options. UK is firmly in the ballot and what in truth is an exclusive excitement of Mt Kenya region and no more. With a supreme registered vote bank & expected awesome turn-out for Muthamaki on the d-day, the "Lesotho" solidarity unfortunately still fall shy of the magic number to clinch the presidency. The ONLY life-line is the HEAVILY anticipated Kalenjin votes that must be delivered in entirety to guarantee victory. Sample this fact; whereas in 2007 Kibaki had 312,000 votes (32%) from the former western province, there is no evidence that Uhuru will achieve anywhere near that figure. That's a deficit indeed and there are more across board. My take is that owing to the "mademoni" saga and all the mis-use and con art on MM & Wamalwa, Uhuru's candidature is poised for a resounding rejection in western with a zeal only a keen to Luo Nyanza. Kibaki harvested the evenly divided Gusii Nyanza votes. This time they're demanding "listening allowance" from a Kenyatta presidential candidate. Jubilee's claim to Kisii & Nyamira votes is a fallacy. Nyanza is looking totally corded and that's another home truth for Raila opponents. Yes Raila lost the Rift-valley mantra, the Rutos, the pentagon, the Shebesh(es) & all the romance of 2007 but he's still standing tall. The Kamba votes were never in his column.. They are an addition in 2013. Albeit by default, CORD is riding a much higher percentage of western into this election than ODM's 07. And finally no one should write off the Kikuyu support for Raila. There are a people from mount Kenya region who could defy the perceived ethnic siege and join Kenyans in voting Raila on merit. All factors equal, Raila-Kalonzo ticket could be encourse to 50+1 win as projected by this polling. Tuko tayari!
|
|
|
Post by Omwenga on Jan 11, 2013 23:59:16 GMT 3
Here is what I blogged about this earlier..., omwenga.com/2013/01/11/poll-predicts-raila-and-cord-will-win-2013-elections/A few days ago I blogged that if the then soon to be released numbers by Infotrak and Synovate Ipsos reflect those that have been conducted privately but not published, those not supporting Raila will be having stomach upsets, triple so those who hate Raila. According to a poll just published by Infotrak, an alliance that brings together Prime Minister Raila Odinga, Vice President Kalonzo Musyoka and Trade Minister Moses Wetangula, Coalition for Reforms and Democracy (CORD) is the most popular coalition at 49 percent, says the Standard, which analyzed the poll data. Deputy Prime Minister Uhuru Kenyatta and MP William Ruto's Jubilee alliance is second in popularity with 40 percent. United Democratic Front led by DPM Musalia Mudavadi comes a distance third with 3 percent in the latest opinion released on Friday. The data shows the most popular presidential aspirant and their running mates are Raila-Kalonzo at 51 percent followed by Uhuru-Ruto at 39 percent, Mudavadi 3 percent, Kenneth-Tuju 3 percent, Karua 0.3 percent and Kiyiapi 0.1 percent. Cord alliance enjoys the most support in North Eastern region with 85 percent followed by Nyanza region at 82 percent while Jubilee alliance is leading in Central region at 85 percent followed by Rift Valley at 52 percent. UDF most support is in Western region at 23 percent. The poll showed that in regions considered to have swing votes, Coast is likely to remain Cord dominant if they increase impetus in their campaigns in the region. Currently they have 58 percent popularity. In Rift Valley, Jubilee will likely remain dominant if the alliance increases its’ impetus in campaigns to protect their popularity. Currently it has 52 percent. I am fairly certain the latest Ipsos Synovate to be published soon will show the same results, if not better for Raila and Cord. Polling by others (not necessarily only those commissioned by Cord) show an even larger lead by Cord well over the 50% +1 margin and very close to crossing the 25 county margin in Round 1 while it will take somewhat of a miracle for Uhuru to cross the 25 county margin even in Round 2, which I maintain there will be not be Round 2. Wait for the usual suspects to come out seething, condemning the pollsters as skewed, biased, blah, blah, blah. Republicans in the US in denial said and did the same thing about polling showing Obama winning against all odds and look who is still crying like babies long after Obama whipped their behinds. The same is happening in Kenya. Every poll taken thus far shows Raila is the man, his opponents and haters refuse to believe but no one is foolish enough to think they don't privately know and are resigned to the fact they can't do anything to stop Raila from being reelected and sworn as our next president. I have been urging people for a long time to start practicing saying HE President Raila Amolo Odinga or be prepared to hear those majestic words for a long time to come, beginning March 4, 2013. Those who dismissed this as wishful thinking, better start doing just that if they haven't already. I fully realize there are those whose heads are buried deep on the ground and wish not to hear or see anything positive about Raila and now Cord, but they too shall soon come to realize all that is in vein destiny is taking us all in one direction and that's where we elect a president and others who have the leadership skills, desire and ability to take our country forward, not backward. There is only one person who can do that at the top among those with any serious chance of being elected and that's none other than Raila Amolo Odinga. That's a fact anyone looking at the field objectively must agree with or disagree with only if being dishonest. Again, practicing those words or getting ready to hear them is advised. Uchambuzi Tanaka, www.omwenga.com
|
|
|
Post by mwalimumkuu on Jan 12, 2013 0:19:36 GMT 3
Infotrak poll concurs with the in-house survey I just carried two days ago, although my findings put Uhuru closer at 45% rather than 39%. There had been a surge of numbers chant over the place. I wanted to get "behind the enemy line" to capture the real situation on the ground. First, the futility of MM's candidature has clearly been exposed and consequently his supposed Luhyia base is overwhelmingly gravitating towards CORD. However he still have remnant support, 25-30% home votes, of which if clings to might curtail RAO's outright win in round 1. Nonetheless in the event of a run-off, all these "displaced" western votes will mostly likey align to RAO, with or without MM's blessing. It goes to say that subject to how he handles the already Messy situation (like his VP pick) Mudavadi's bid could metamorphosize into a total joke where by come the ballot day his candidature will be "pro non scripto" ie negligible even among his Luhyia diehards. That is when he will finish with a Shikuku/Kukubo votes as opposed to Kijana Wamalwa's (1997) that he's currently envying for the sake of a bargain.. In the aforesaid scenario, CORD becomes a unanimous alternative of western votes; a decisive support that would most certainly put Raila home & dry in 90 minutes. Second notable issue about this election is the Kalenjin's Rift Valley votes ( registered at 1.5M). The jubilee (read TNA ) numbers' bravado is HEAVILY pegged on the assumption that this vote will be herded in-toto to face mt Kenya; because Ruto is VP and the "promise" of 50% govt share...Oh yes a promise, a binding MoU! This is where the biggest surprise awaits the dark forces. As a fact on the ground, the ODM/ Arap Mibei wave that swept the Kalenjin political landscape in 2007 will neither be swapped nor be replicated this election cycle. Equally, although no one is discounting the existence of Uhuru's support among the Kalenjin, there are some known unknowns here! If I'm in jubilee, I wouldn't toss a champaigne yet over Rift valley votes beyond Nakuru & Laikipia. Achicha Wendi Kityo is a brewing undercurrent that could spoil the mountain party. My statistical observation is that 30% Kalenjin vote evading Mr Kenyatta will leave him in deep hole. Apportion that number to Raila and the party could be over. This is only the Kalenjin constituency without factoring the Maasai, Samburus & the Turkanas of Rift-Valley. In other words a modest support from Ruto's stronghold is just enough for CORD riding forward given the overall picture of the race. Mambo bado! ABR (anybodybutraila) scheme through MM & western has seriously boomeranged. It's too little too late to remedy that front. UK is firmly in the ballot and what in truth is an exclusive excitement of Mt Kenya region and no more. With a supreme registered vote bank & expected awesome turn-out for Muthamaki on the d-day, the "Lesotho" solidarity unfortunately still fall shy of the magic number to clinch the presidency. The only life-line is the Kalenjin votes that must be delivered in entirety as top up to victory. Sample this fact; whereas in 2007 Kibaki had 312,000 votes (32%) from the former western province, there is no evidence that Uhuru will post anything close that figure especially after the "mademoni" trick on MM. That's a deficit and there is more of that across the board. Kibaki harvested the evenly divided Gusii Nyanza votes. This time they're demanding "listening allowance" from a Kenyatta presidential candidate. Jubilee's claim to Kisii & Nyamira votes is a fallacy. Nyanza is looking totally corded and that's another home truth for Raila opponents. Yes Raila lost the Rift-valley mantra, the Rutos, the pentagon, the Shebesh(es) & all the romance of 2007 but he's still standing tall. The Kamba votes were never in his column.. They are an addition in 2013. Albeit by default, CORD is riding a much higher percentage of western into this election than ODM's 07. And finally no one should write off the Kikuyu support for Raila. There are a people from mount Kenya region who could defy the perceived ethnic siege and join Kenyans in voting Raila on merit. All factors equal, Raila-Kalonzo ticket could be encourse to 50+1 win as projected by this polling. Tuko tayari! Sam Ochola,I see where you are coming from. Whereas your argument makes more sense than the so called scientific poll, you can see clearly that the road to victory for CORD is dependent on things going very wrong for Jubilee in both RV and Central, which is a very long shot in my own estimation. If central can resist Mudavadi the way they just did, what would make them vote RAO? As long as Uhuru is on the ballot, any candidate cheating themselves that they can get anything more than 1% in central need their heads examined. It is a dream, a very bad dream. Secondly, the rage in RV against Raila is real and more visible than against Kikuyus currently. I was in Nandi county the other day and was shocked at the lack of ODM presence in that county despite being the home of ODM chair. Even him (Koskei), people are not yet sure whether he is running or not, because there is nothing on the ground to show where he is headed. Down in Emgwen constituency, out of 9 candidates, 7 are URP, 1 UDF and 1 KANU, no ODM, and the same scenario is replicated in the other neighboring constituencies. As regards the Eastern vote, I totally agree with one of the contributors here (I cant remember who) who argued that, Kalonzo effect will not be as huge as it was in 07, simply because he is not running for president. Moreover, there is a huge Kamba population that does not just trust Raila at all, especially given the treatment Kalonzo continued to receive from the ODM leadership throughout the life of this coalition government. Couple that with the fact that Kalonzo himself is a very weak politician without any record in Ukambani, and that he is up against one of the very few very good mobilizers in our politics in the name of Ngilu. I also think your dismissing of Mudavadi's candidature is premature. Remember he just begun his maiden tour of the region as the candidate designate for Amani Coalition. We need to wait and see how that will unfold. The fact that CORD elected to visit the region the same time he is in the area tells me, there is panic in RAO corner and they would want to stem huge movement of the electorate to the Amani side. One thing for sure, with Wamalwa in his corner, he now has a very solid bridge to the Bungoma and majority of the Trans Nzoia vote, which he did not have before. The love of the people there for the Wamalwas is quite significant. A lot is also dependent on his choice of running mate and how he manages his campaigns. So far he is looking very good in Vihiga and Kakamega counties. If he can maintain those figures, penetrate Bungoma and Busia, and show a lot of energy and optimism in his campaigns he will be a big headache to CORD.
|
|
|
Post by Sam Ochola on Jan 12, 2013 8:34:38 GMT 3
I noticed there is the usual dispute on how scientific these polls are. I appeal to those who are meticulous on methodology and variable definition thus: Help me out folks: Over the last six months I have reached the conclusion that, whereas they are other variables influencing the choice of PORK, ETHNICITY will be { outside of Nairobi} the determinant factor amongst the big ethnic groups the GEMA, the Luo, the Luhya, and The Kalenjin. I therefore miss an opinion poll which breaks it down for me: 1. What percentage of Kikuyu voters will vote Jubilee? 2. What percentage of Jaluos will vote Cord? and so too Kalenjin and Luhya. I did not study mathematics, but I think ETHNICITY may be, in this case in Kenya of the March 4 election, a better scientific indicator of the voting preference, than the other ' statistically representative Kenyan'. I doubt there is anything definable as Kenyan in you mates! that is worth a scientific variable: that would be a holy grail tale! The representative Kenya can then be confined to the minority tribes who give no damn anyway, coz they have no hope in hell of getting their man to the top job. Consequentially with a low turn out. Now, has any Jukwaaist given a dense amount of thought to using ethnicity as a scientific yard to predict party allegiance? [That is what zoning is all about anyway, but do we say!] STATE OF KENYA'S PRESIDENTIAL ELECTION, PROJECTED SURVEY OF THE POLL OUTCOME, JANUARY 10TH 2013. BY SAM OCHOLA NO, REGIONS, REGISTERED VOTER S, PROJECTED TURN-OUT, PROJECTED CAST VOTES, CORD % SHARE, CORD VOTE SHARE, JUBILEE % SHARE, JUBILEE VOTE SHARE, AMANI & OTHERS %, AMANI & OTHERS VOTE SHARE 1 COAST PROVINCE 1164083 57% 663527 60% 398116 38% 252140 2.00% 13271 2 NORTHERN FRONTIER 504482 50% 252241 60% 151345 38% 95852 2.00% 5045 3 MERU/EMBU 866329 70% 606430 30% 181929 68% 412373 2.00% 12129 4 UKAMBANI 1063529 65% 691294 85% 587600 13% 89868 2.00% 13826 5 CENTRAL PROVINCE 2190477 80% 1752382 8% 140191 90% 1577143 2.00% 35048 6 TURKANA/SAMBURU/POKOT 284901 60% 170941 60% 102564 38% 64957 2.00% 3419 7 NAROK/KAJIADO 568139 69% 392016 60% 235210 38% 148966 2.00% 7840 8 NAKURU/LAIKIPIA 866146 75% 649610 30% 194883 68% 441734 2.00% 12992 9 KALENJIN RIFT VALLEY 1556060 80% 1244848 33% 410800 65% 809151 2.00% 24897 10 BUNGOMA REGION 510588 65% 331882 80% 265506 8% 26551 12.00% 39826 11 KAKAMEGA COUNTY 568,813 65% 369728 75% 277296 5% 18486 20.00% 73946 12 BUSIA COUNTY 251737 65% 163629 85% 139085 3% 4909 12.00% 19635 13 VIHIGA COUNTY 202456 65% 131596 70% 92117 2% 2632 28.00% 36847 14 LUO NYANZA 1257849 80% 1006279 97% 976091 1% 10063 2.00% 20126 15 GUSII/KURIA NYANZA 696907 65% 452990 68% 308033 30% 135897 2.00% 9060 16 NAIROBI 1778903 70% 1245232 50% 622616 47% 585259 3.00% 37357 p NATIONAL TALLY 14331399 71% 10124625 50% 5083381 46% 4675982 4% 365262 VICTORY MARGIN (% AND VOTE SPREAD BETWEEN CORD & JUBILEE) 4% 407,399 APPENDIX 1 NORTHERN FRONTIER: North Eastern province; plus Isiolo & Marsabit Counties (ie Somali/pastoralists) 2 KALENJIN RIFT VALLEY: Nandi, Uasin-Gishu, Elgeyo-Marakwet, Baringo, Kericho & Bomet Counties; plus Mt Elgon, Endebess & Cherangany constituencies. 3 BUNGOMA REGION: Bungoma & Tranzoia Counties; excluding Mt Elgon, Endebess & Cherangany constituencies 4 TURKANA/SAMBURU/POKOT: Turkana, Samburu & West Pokot Counties. 5 GUSII/KURIA NYANZA: Kisii & Nyamira Counties; including Kuria East & Kuria West constituencies. 6 NAROK/KAJIADO: Narok & Kajiado Counties (ie Maasailand) 7 MERU/EMBU: Meru, Tharaka & Embu counties. (ie Mt Kenya East) 8 UKAMBANI: Machakos, Makueni & Kitui Counties (ie Lower Eastern). 9 NAKURU/LAIKIPIA: Nakuru & Laikipia Counties (ie Central Rift Valley) 10 Primary Data Source: IEBC BVR 18th Dec 2012 11 I have deduced Regional and national voter turn-out on account of past elections ( eg 71% in 2007 & 69% 2010 referendum), historical trends & current ethno-political dispensation. The net has yielded 71% turn out for March 4th 2013. Not surprising, the turn-out in Nairobi alligns with the national tally; a concurrence with history. 12 Kikuyu, Luo & Kalenjin areas have history of the most proactive voter participation. Borrowing closely frm recent polls (2007 & 2010), I have apportioned 80% voter turn-out in the 3 regions. for the purpose of my analysis. The voter enthusiasm that greeted ODM wave in Kalenjin 2007 will however not be replicated in 2013 by any side. Thus 80% Kalenjin turn out is an ambitious projection. 13 A much lower turn-out, 50-57% is apportioned to Coast, Northern frontier & Turkana/Pokot areas to account for probable insecurity challenges, MRC insurrection & associated hardships/ apathy. In 2007, Raila posted 59%, Kibaki 33% and Kalonzo 6%. There is no reason to discount Cord's support at 60% in the region. The presence of Naomi Shaban, Mwakwere, Mungutana and Balala's in TNA does not translate to a sweep of the region. These politicians have their price tags and Uhuru's money simply purchased them. Land, devolution and resource grievance runs deep in places like Taita-Taveta 14 A stark reminder of the issues and Uhuru Kenyatta's candidancy becomes a hard-sell. I don't need further explanations. I've allocated Jubilee 38% vote share in Coast, reasonable figure. For similar land issues, Uhuru will meet stiff resistance from Narok & Kajiado Counties when push comes to shove (don't forget Ole NtimamaS are still in town). There is however significant Jubilee support in the two counties notably the Kikuyu Diaspora settled here. I have thus apportioned 60-38% vote share in CORD's favour within Maasailand 15 Track record in Ukambani, Luhyia & Kisii indicate equivocal & lustre voter participation hence the conservative turn-out projection of 65% in these regions. This is below the national average of 70%. My take is there is no Nyachae; Kalonzo is not on the top ballot & the "mademoni" made Mudavadi's prospects so slim as warrant the effort. A note of caution: protest vote (don't ask me against who) could inspire solid turn-out in both Ukambani & western to 70% +. The scenario would be a game changer in favour of CORD but remains to be seen. 16 Unless the Kalenjin RV (as in my table) vote " Uhuru for President" with the same propensity of "Arap Mibei" in 2007, Mr. Kenyatta cannot NOT attain 50+1. regardless of all efforts elsewhere (even if gets 100% from central). I stand sober with 65% vote share for Uhuru from this region. URP will definitely sweep other elective positions but Presidential return for Uhuru amongst the Kalenjin could only peak to 70%. That's the feeling on the ground. Although Gusii/Kuria have traditionally steered off the Nyanza voting pattern, the trend on the ground inclines towards inevitable solidarity with Luo counterparts for CORD presidential ticket. Anyone keen to observe the recent Uhuruto visit will attest to this. The most UK can achieve in Gusii counties is 30% vote share. 17 Raila needs a third of Kalenjin votes and some 25% of Meru/Embu vote basket to be encourse for 50+1 in first round. Both are modest margins within his reach. (I have accordingly reversed the postings of 2007 and allocated Raila 33% & 65% for UK in Rift Valley. This is a safe projection)
|
|
|
Post by mank on Jan 12, 2013 8:41:31 GMT 3
Omwenga, Most of the discussion is actually about the alleged scientific method of this study ... its not about Raila.
|
|
|
Post by kamalet on Jan 12, 2013 11:43:50 GMT 3
Faulting the Ochola predictions.
NEP in my view is a toss up and will turn out at 50-50. It is wrong to include Isiolo and Marsabit as these are essentially what we call upper Eastern.having said this we can give it all to CORD.
In the Meru/Embu figures,unless Ochola sees North Imenti votes totaling up to 181,000 votes on account of Imanyara's support for CORD the estimates ate fatally wrong. I would call it 85% for Jubilee and 10% for CORD.
In Ukambani the figures are wrong apart from the recognition that Kalonzo as DPORK will not attract that many votes as he did in 2007. Having said that it is holy unlikely that disgruntled Kambas and supporters of Ngilu only add up to 90,000 votes. I would actually call it 35% for Jubilee and the balance to CORD.
I could concede to the Turkana/Samburu/Pokot as per Ochola though my view is that it is perhaps the Turkana only who are leaning the Cord side but leave that as it is.
The Narok/Kajiado figure are actually inverted and I would reverse therefor Jubilee based on the support that TNA has in Kajiado and the URP support in Narok.
As for Nakuru and Laikipia votes just how does CORD get nearly 200,000 votes in a virtually kikuyu constituency? From 30% I would give Cord 10%.
As the flaws continue I agree with Luo Nyanza voting 97% for Cord, but where is Cord getting 8% from unless it is the padding process!
As regards Luhyia country, I think Ochola has ignored Mudavadi and that will be at CORDs peril. I would be looking at 60% Cord and the balance to Mudavadi if the man remains lazy and fails to consolidate the community vote which he might if he works hard.
When the revisions are done I am still getting 53% for Jubilee and 45% for CORD and even worse depending on how Mudavadi hacks it. The clearest position of how this will concretize is next Saturday after the nominations.
|
|
|
Post by phil on Jan 12, 2013 16:47:11 GMT 3
in response to Alex Kuria who asked Que.why did Infotrak Ltd only carry out the survey in 25 counties?
Good Question. When designing a sampling frame, one is guided by available population statistics that represent the target population. In our case, the target population is Registered Kenyan Voters who happen to reside all parts of the country .The following administrative units can be used to guide the sampling frame:
1. The 8 regions formerly known as provinces
2. The 17 electoral regions
3. The 47 counties
4. The 290 constituencies
We use the 8 regions as our first guideline. Here we ensure that we cover all the regions. We then look at the 47 counties and the constituencies and wards within the respective counties. Now, you don’t have to sample ALL the counties for the study to be nationally representative. In each region we purposively select counties that are representative of the different people who reside in the regions. Given that the sample is 1500, it would be foolhardy to cover all the counties. It would mean covering very small samples in each county. When we analyse the numbers, we are keen to ensure that each unit of analysis is significant for us to make useful inferences. In other words 20 or even 25 objectively selected counties is more advisable that very small samples in each of the counties. Angela.
|
|
|
Post by OtishOtish on Jan 12, 2013 17:39:36 GMT 3
Ochola:
Interesting posts but hard to read in the entirety. Whenever you are ready to, feel free to use paragraphs. ;D
|
|
|
Post by foresight on Jan 12, 2013 17:54:50 GMT 3
Raila Amolo Odinga will be next president of the republic of Kenya. I can also prove it mathematically why Uhuru Kenyatta won't. Before I get complex just know that as we stand only Rift Valley and Central province are sure bets for Jubilee. And you will be pleasantly surprised by the huge votes CORD WILL GET in RV. Eastern, Western, Coast, Nyanza, Nairobi are all leaning heavily towards CORD. Then this HAGUE thing will not go away and it's getting really serious as we speak, jubilee leaders have reasons to worry, also no sane foreign power will deal with the Jubilee leaders the day they step at the Hague doors will slam shut on their faces. They will be hard pressed to visit even Sudan. The poor leadership skills we have witnessed lately from jubilee has really opened the eyes of many people and as such JUbilee has been on a path of self destruction in terms of voter appeal outside their strongholds. Just watch the fighting that will take place within that alliance during the nominations. One need not be a prophet to predict the obvious.
|
|
|
Post by reporter911 on Jan 12, 2013 18:49:05 GMT 3
Raila Amolo Odinga will be next president of the republic of Kenya. I can also prove it mathematically why Uhuru Kenyatta won't. Before I get complex just know that as we stand only Rift Valley and Central province are sure bets for Jubilee. And you will be pleasantly surprised by the huge votes CORD WILL GET in RV. Eastern, Western, Coast, Nyanza, Nairobi are all leaning heavily towards CORD. Then this HAGUE thing will not go away and it's getting really serious as we speak, jubilee leaders have reasons to worry, also no sane foreign power will deal with the Jubilee leaders the day they step at the Hague doors will slam shut on their faces. They will be hard pressed to visit even Sudan. The poor leadership skills we have witnessed lately from jubilee has really opened the eyes of many people and as such JUbilee has been on a path of self destruction in terms of voter appeal outside their strongholds. Just watch the fighting that will take place within that alliance during the nominations. One need not be a prophet to predict the obvious. HERE IS FOURTH PRESIDENT OF KENYA- WAPENDE NA WASIPENDE!!Hey let those who are in denial like Kamalet, joblesscorner, mank, Mwalimu, have their weak voices heard, in the sea of the coming "CORD" Tsunami heading for a landslide "WIN" the only thing Kenyans have to watch out for is the rigging using the government machinery to swear their protégé in at night.. ( 2007 taught Kenyans a lesson, 2013 Kenyans are prepared and vigilant) it ain't going to happen again..
|
|
|
Post by mangai on Jan 12, 2013 18:58:18 GMT 3
Before we crucify the messenger, in this case Infotrack, please have a look at opinion polls done recently by Ipsos/Synovate (Uhuru leaning) and Strategic Research (or is it another Raila leaning?). Take note that these two were done before Mudavadi was ejected from Jubilee.
Looking carefully at some of the counties sampled by Infotrack, Nandi County was for instance polled at 94% in favour of Jubilee which agrees with some sentiments expressed herein.
Be the Judge!
14th December 2012-Ipsos/Synovate Opinion Poll
The latest Ipsos/Synovate opinion poll holds some good news for Raila Odinga, his running mate Kalonzo Musyoka and their Coalitions of Reforms and Democracy (CORD) but the figures suggest it is all still to play for in the race for Kenya’s presidency with under three months to go to polling day.
ODINGA-MUSYOKA vs KENYATTA/RUTO The Ipsos poll carried by phone across 1,625 respondents between December 6 and 11 suggests that if the election had been held at the time the Odinga/Musyoka ticket would have garnered 47 per cent of the vote against Uhuru Kenyatta and William Ruto’s 41 per cent, with 12 per cent of the electorate ‘undecided’.
With a margin of error in the poll of +/- 2.43 per cent and more than 1-in-10 voters yet to make up their mind as to who they will vote for, the result is still very much in the balance and it still looks more than probable that the election will go to two rounds as no one candidate seems likely to reach the 50 per cent + 1 threshold required for victory in the first round.
ODINGA-MUSYOKA vs RUTO-MUDAVADI Should the current candidate alliances change for whatever reason, the figures still look better for Odinga and Kalonzo. Pitched against Ruto and Musalia Mudavadi, for example, the Ipsos poll suggests Raila’s team would win the first round of the election by 48 per cent to 38 per cent, with 14 per cent of voters remaining ‘undecided’ on that alignment of candidates.
21st December 2012-Strategic Africa
Cord trounces Jubilee in new opinion poll Prime Minister Raila Odinga’s Cord has topped the latest opinion poll as the most preferred coalition in the country.
The poll, conducted by Strategic Africa, shows that Cord is the most popular alliance in the country with popularity of 49.7%. Jubilee, despite its much publicised wrangles, polled second at 43%.
Peter Kenneth and Raphael Tuju’s coalition came in third at 3.1% while Eugene Wamalwa’s Pambazuka alliance polled at 1.7 %.
When respondents were asked who they preferred as their coalitions flag bearer, 92.8% of the respondents stated they would prefer Raila Odinga as Cord’s flag bearer while 6.2% prefer Kalonzo Musyoka.
In the Jubilee alliance which comprises Uhuru, Mudavadi And Ruto, 71.9% would prefer Uhuru Kenyatta as the presidential candidate whereas 18.4% would like to see Musalia Mudavadi as the Jubilee flag bearer.
The poll which was released on Friday, agrees with Ipsos Synovate’s recent survey which showed in a possible scenario where Raila is twinned with Kalonzo and pitted against an Uhuru/Ruto ticket, 47 per cent of Kenyans would vote Raila/Kalonzo while 41 per cent would vote for the later.
|
|
|
Post by mank on Jan 13, 2013 21:12:34 GMT 3
in response to Alex Kuria who asked Que.why did Infotrak Ltd only carry out the survey in 25 counties?
Good Question. When designing a sampling frame, one is guided by available population statistics that represent the target population. In our case, the target population is Registered Kenyan Voters who happen to reside all parts of the country .The following administrative units can be used to guide the sampling frame:
1. The 8 regions formerly known as provinces
2. The 17 electoral regions
3. The 47 counties
4. The 290 constituencies
We use the 8 regions as our first guideline. Here we ensure that we cover all the regions. We then look at the 47 counties and the constituencies and wards within the respective counties. Now, you don’t have to sample ALL the counties for the study to be nationally representative. In each region we purposively select counties that are representative of the different people who reside in the regions. Given that the sample is 1500, it would be foolhardy to cover all the counties. It would mean covering very small samples in each county. When we analyse the numbers, we are keen to ensure that each unit of analysis is significant for us to make useful inferences. In other words 20 or even 25 objectively selected counties is more advisable that very small samples in each of the counties. Angela. That seems to address the methodological concerns. However the practicability of these sentiments in context of the subject of study is suspect. What considerations went into the identification of counties that are representative of the different people residing in a region? If we think about it, counties are just former districts, and for most parts those are perfectly tribal enclaves. So, how can tribal enclaves be representative of of the different people that leave in a region of several tribal enclaves?
|
|
|
Post by nowayhaha on Jan 13, 2013 21:17:39 GMT 3
my suggestion to pollsters. 1) Use registered votes..Infotrak used 18th dec..but if you compare their data with IEBC data..its INACCURATE..they claim Nairobi has 2.1M registered votes...whilst we know 1.7M registered..Did they cross check that?. In RV they claim 3.1M..whilst we know 3.4M registered... With just that single error...INFOTRAK has lost neary 0.8M..about 6-7%!!!!!!!!!!!!!!!!!! of their result therefore become erroneous. You cannot call that scientific. Error margins of +-6% ignoring data collection errors, data entry errors, sampling errors. 2) Dump province unless you know how to navigate RV, Eastern, Coast, Western and Nyanza. RV is very complicated. And it's 25% of the vote. You get it wrong...and you miss the plot. You can use IEBC regions or you have to JUDICIOUSLY select counties that are representative. Infotrak purposive selection in RV and Eastern are erroneous. Get another strata..forget about province.. Purposive criteria SHOULD BE TRIBE. You cannot claim to have possessively selected counties without any justification. 3) Do proper household survey..not telephone calls..randomly select folks..make sure they show ID card..show voter cards. Finally have high quality data handling and analysis from get go. Infotrak if they cannot notice Nairobi 2.1M vote ERROR cannot possible produce high quality poll you would expect of a scientific poll.
|
|
|
Post by reporter911 on Jan 14, 2013 2:55:53 GMT 3
my suggestion to pollsters. 1) Use registered votes..Infotrak used 18th dec..but if you compare their data with IEBC data..its INACCURATE..they claim Nairobi has 2.1M registered votes...whilst we know 1.7M registered..Did they cross check that?. In RV they claim 3.1M..whilst we know 3.4M registered... With just that single error...INFOTRAK has lost neary 0.8M..about 6-7%!!!!!!!!!!!!!!!!!! of their result therefore become erroneous. You cannot call that scientific. Error margins of +-6% ignoring data collection errors, data entry errors, sampling errors. 2) Dump province unless you know how to navigate RV, Eastern, Coast, Western and Nyanza. RV is very complicated. And it's 25% of the vote. You get it wrong...and you miss the plot. You can use IEBC regions or you have to JUDICIOUSLY select counties that are representative. Infotrak purposive selection in RV and Eastern are erroneous. Get another strata..forget about province.. Purposive criteria SHOULD BE TRIBE. You cannot claim to have possessively selected counties without any justification. 3) Do proper household survey..not telephone calls..randomly select folks..make sure they show ID card..show voter cards. Finally have high quality data handling and analysis from get go. Infotrak if they cannot notice Nairobi 2.1M vote ERROR cannot possible produce high quality poll you would expect of a scientific poll. feast on these polls until the next ones come out showing "CORD" Tsunami.. ;D ;D
|
|
|
Post by jakaswanga on Jan 14, 2013 20:51:33 GMT 3
I noticed there is the usual dispute on how scientific these polls are.
I appeal to those who are meticulous on methodology and variable definition thus: Help me out folks: Over the last six months I have reached the conclusion that, whereas they are other variables influencing the choice of PORK, ETHNICITY will be {outside of Nairobi} the determinant factor amongst the big ethnic groups the GEMA, the Luo, the Luhya, and The Kalenjin.
I therefore miss an opinion poll which breaks it down for me: 1. What percentage of Kikuyu voters will vote Jubilee? 2. What percentage of Jaluos will vote Cord? and so too Kalenjin and Luhya.
I did not study mathematics, but I think ETHNICITY may be, in this case in Kenya of the March 4 election, a better scientific indicator of the voting preference, than the other 'statistically representative Kenyan'.
I doubt there is anything definable as Kenyan in you mates! that is worth a scientific variable: that would be a holy grail tale!
The representative Kenya can then be confined to the minority tribes who give no damn anyway, coz they have no hope in hell of getting their man to the top job. Consequentially with a low turn out.
Now, has any Jukwaaist given a dense amount of thought to using ethnicity as a scientific yard to predict party allegiance? [That is what zoning is all about anyway, but do we say!] Even in the study design you are proposing we would still want to see a more encompassing casting of the sampliing net. And you are right, ethinicity is the most important defining factor of the typical Kenyan's voting inclinition. The simple static design mostly used by statisticians (like the topical one) is premised on the idea that simple stastical measures (mean, standard deviation and the like) of outcomes in a representative sample are transferrable to the population (key qualifier: representative sample). This premise is that if your sample reflects the population, then you do not need to isolate the individual effects of the variables defining population behaviour .... you study the sample as if it is a full cross section of the population. so it is obvious what happens when you decide that a certain ethinicity (e.g. Embu) need not be sampled (for Eastern Province). The method you are proposing is more ambitious: it seeks to isolate the individual effects of various explanatory variables for the outcome of interest, and then to apply those effects to the defining variables of the population to come up with population effect. That's the mainstay of economists (econometrics). Either way, the representativeness of a sample is important, in that if you miss to sample enough study units with an important determining factor, your results will be suggesting that such a factor is unimportant. ... but the second method, is less dependent on exact representativeness of the population ( i.e. it does not asume that the sample is like full crossection of the popluation, rather it assumes that all the important decision variables of the population are sufficiently represented in the sample) - it requires, of course, that the sample elements on average look like those of the population (i.e. you are not sampling only the extremes - and it turns out that random large samples will generally furnish this quality), and that there be enough study units with expression in each of the variables of importance (you might have had of some jargon like degrees of freedom, system identification, etc - all those come into play here) so that individual variable effects can be mathematically extracted. In this case the authors choose to use the static method which requires sample likeness with population, and then they do not include all ethinicities (yet we know the importance of every ethnicity). When studies do this, often they make the assumption that some ethnicities are like copy cats of others, e.g. they might say that once you have sampled the Meru you do not need to sample the Embu. That might work in certain cases ... not so elections. In fact even Meru by itself is very polarized, and if sampling in Meru does not balance that polarity it will end up biasing the result, suggesting that Meru is more like the locality it studies more ... and in this case, it will extend the same biase to Embu and to Eastern province in general. Mank, Econometrics it be yes. First I want to state this caveat openly, before I say my piece. I am aware there have been controversies of a very serious nature in the USA and beyond, whenever research has included ethnicity as a parameter. Eg, in the supposed relative intelligence of races. There was also the matter of eugenics, claimed as a science and used by THE NAZI state of Germany to Holocaustic effect. Ethnic profiling, for whatever purpose, I aware then, is a mine-field, and I would only too gladly ignore it myself, were it not that I believe, and think, ignoring it in the case of the coming Kenyan elections is a bit the old saying of the ostrich in a particular position. ---So said. What I wanted to think about, Mank, with respect to econometric modules, is this: after we have established BIAS respective ethnic group, EG that 90% of Luos will vote CORD and 90% of Gikuyus Jubilee, I think we can dismiss any member of the above group as a representative Kenyan. But we can mathematically work out what a representative Luo is, and what a representative Gikuyu is, and denote them thus. And it appears to me, it would be very complex math to whitewash a representative Gikuyu or Luo into a representative Kenyan. That is, at over 90% loyalty to specific [opposing] parties, these two groups are statistically PREJUDICED. So I would propose every [Major] Ethnic Group in Kenya [they will total 68% of the electorate I think] be polled apart, so that their BIAS of PREJUDICE factor, inclination, be noted apart. This, I think, sheds a clearer light on their actual preference come the voting day. Then the much more neutral groups can be polled on their own. ---The math to arrive at the credible poll gets complex yes, but that is the econometrics then! Problem I think, is this method would be more expensive and time consuming. So lets say, come March 14, history shall absolve Infotrak, or Harris, or Gallup, or Synovate. And their methodology emerge victorious in the test of reality. Their reputation intact. Good for them, but does not mean the model can not be fine-tuned. [In the USA I think they have some ethnic and sex tailored polls too. So I gathered Obama rated a whooping 98% under ALL categories black women. And Romney 95% amongst white MALES above 70 years! NB: Mank, have we discussed 'representative Kenyan' before? and % swings? embedded somewhere in Jukwaa archives? I seem to think so!
|
|
|
Post by mank on Jan 15, 2013 7:40:20 GMT 3
Mank,
Econometrics it be yes.
First I want to state this caveat openly, before I say my piece. I am aware there have been controversies of a very serious nature in the USA and beyond, whenever research has included ethnicity as a parameter. Eg, in the supposed relative intelligence of races. There was also the matter of eugenics, claimed as a science and used by THE NAZI state of Germany to Holocaustic effect. Ethnic profiling, for whatever purpose, I aware then, is a mine-field, and I would only too gladly ignore it myself, were it not that I believe, and think, ignoring it in the case of the coming Kenyan elections is a bit the old saying of the ostrich in a particular position. ---So said.
What I wanted to think about, Mank, with respect to econometric modules, is this: after we have established BIAS respective ethnic group, EG that 90% of Luos will vote CORD and 90% of Gikuyus Jubilee, I think we can dismiss any member of the above group as a representative Kenyan. But we can mathematically work out what a representative Luo is, and what a representative Gikuyu is, and denote them thus. And it appears to me, it would be very complex math to whitewash a representative Gikuyu or Luo into a representative Kenyan. That is, at over 90% loyalty to specific [opposing] parties, these two groups are statistically PREJUDICED.
So I would propose every [Major] Ethnic Group in Kenya [they will total 68% of the electorate I think] be polled apart, so that their BIAS of PREJUDICE factor, inclination, be noted apart. This, I think, sheds a clearer light on their actual preference come the voting day. Then the much more neutral groups can be polled on their own. ---The math to arrive at the credible poll gets complex yes, but that is the econometrics then!
Problem I think, is this method would be more expensive and time consuming. So lets say, come March 14, history shall absolve Infotrak, or Harris, or Gallup, or Synovate. And their methodology emerge victorious in the test of reality. Their reputation intact.
Good for them, but does not mean the model can not be fine-tuned. [In the USA I think they have some ethnic and sex tailored polls too. So I gathered Obama rated a whooping 98% under ALL categories black women. And Romney 95% amongst white MALES above 70 years!
NB: Mank, have we discussed 'representative Kenyan' before? and % swings? embedded somewhere in Jukwaa archives? I seem to think so! Jakaswanga, You are a real trouble maker, you! You already pushed this discussion way beyond social blog material levels, and you are pushing it several notches higher again. I am not sure if I should take your bait and go as far as your teasing seems to ask - but you have my econometric taste buds already sensitized. In econometric analysis we are usually not worried about the kind of controversy you highlight - instead we are concerned about the objectivity of our model definition. Economists try to understand phenomena from a totally objective stance; as such, in this case, since you and I determine that an ethnic variable is necessary in a voting model, we would not be motivated by worries about the reader's sensitivities to the extent of dropping such a variable from our model. We would put it in there, and perhaps discuss those sensitivities in our policy implications or other relevant part of our report (could call it "Caveats", stating any caution we deem fit to disarm the sensitive reader from misusing our study. If we found it too risky to include an ethnic variable or its proxy, then we probably would not undertake the study. Of course if a variable is sensitive, there might be no data for it ... in such a case we might be forced to use a weak proxy for that reason (data challenges). On the other hand, I have found many behavioral scientists creating their own effects through the kind of models they define - too often they plant non-existent evidence into their results, coined from their theories even when the data does not speak to such outcomes. Too often they twist data and plant their own effects into it in the process. We discussed one such study here on Jukwaa (some respect to man Einstein in that regard). So there is good reason why certain variables attract unfair attention. However when you define variables objectively and let the data speak for itself, then you have nothing to worry about - you are simply eliciting an existing effect (bringing to the open "what is"). On The Typical Kenyan Question: In the field of econometrics we could (and should for this type of question) impute the typical Kenyan. If we think only about the ethnicity of that amorphous Kenyan, the person will be more Kikuyu (22%) than any of the rest of tribes. It will be slightly less Luhya(14%), and even slightly less Luo (13%) ... that Kenyan will be negligibly El Molo (2012 demographic profile). You already see how we are doing this creation. On Gender, that Kenyan will be almost 50:50 male:female per population statistics (albeit slightly more male). For an election model we would like to define these attributes from the voting age profile of Kenyans. What we are doing in creating the typical Kenyan is simply melting the characteristics of the population into dough, then from that dough we are molding a single person in the likeness of Kenya's population. Some people may wonder what part of this person will be Kikuyu and what parts will be Luo, but I don't think you will be one of those . Once we have created this Kenyan, we send it to vote (in our model) ... the result is our forecast of the election. There are a number of models that could be used to elicit the voting outcome of a multivariate decision model like the one we are discussing. LOGIT and PROBIT models will define the voting preference against the characteristics of the voter (which in this case would include Ethnicity). Generally the model will be like P(Raila, Uhuru, other contestants) = g(X), where P could be a probabilistic (Probit models)or a Logistic (Logit models) function. Solving this model (don't ask how we do that), given the data and a good identification of the X matrix of voting determinants, then we shall parametrize the prospects of Raila, Uhuru and whoever else we think is worth the pain, against the telling variables. ... from there we simply subject our parametrized model to the characteristics of that unenviable typical Kenyan we created from demographic data of the voting profile. A maximum Likelihood Estimation model could also be defined for the problem. On the subject of the Infotrak study being absolved by history come March 14, I would say we should be cautious in judgement. The fact is that Infotracks winner could actually be the winner even while the model is wrong. Based on the limited info I have read about the model, I don't think the model is very realistic - yet we know Raila has good prospects (just telling from common noise). So I doubt I would change my view on the quality appropriateness of the study even if the election result aligns with the forcasted results .... except of course I become aware of favourable details that I now miss. Indeed we have discussed something of the sort (I recall particularly the discussion of %swings).
|
|