Friday, January 05, 2007

Steadman Polls

There has been a heated debate on the accuracy of Steadman Polls. Those who feel cheated are rubbishing it as poll financed by Colonialists and the Kibaki administration. Critics of the polls are not questioning the methodology used to conduct the polls. Their beef is that the results don’t make sense to them because they believe their candidate should be the next president.

These individuals use a classical fallacy of composition to validate their opinion. Sample this:

“Kenyans reported approval of Kibaki’s record is suspect. Have they forgotten his unfulfilled pledges, corruption, Arturs’ saga and the killing of civilians by police? Or what is Steadman up to?” http://www.eastandard.net/archives/cl/hm_news/news_s.php?articleid=1143963304&date=4/1/2007

Miguna’s argument proclaims that if Kenyans are angry about the things he listed, then they must disapprove Kibaki with a vote of 100%. He also assumes that if he disapproves Kibaki because of the reasons he listed then all Kenyans must disapprove him on those grounds.

Miguna argues: “Take, for instance, the reported results of how Kenyans view their political parties. The first error the Steadman poll committed was to attempt to compare Narc, which is a coalition of more than ten political parties (including LDP), with LDP … if this question was to be considered valid, then people’s views should also have been sought of Narc versus DP; Narc versus Ford Kenya; Narc versus Ford People; Narc versus Safina; Narc versus Sisi kwa Sisi; Narc versus Kanu.” http://www.eastandard.net/archives/cl/print/news.php?articleid=33993

One point that Miguna ignores is that LDP does not count itself as part of Narc. LDP has been presenting herself as an opponent of Narc. I bet Miguna will be throwing a tantrum if LDP was not presented for polling.

Miguna goes on to say: “But there is something even more shocking in the Steadman poll claim that needs to be exposed. The most favoured presidential candidate is Kalonzo Musyoka of the LDP with about 35 percent. This is not unreasonable when we consider the popularity of LDP and ODM. What is shocking and completely unbelievable is the claim that one of the most recognised and leading lights in the LDP and ODM, Raila Odinga, did not even qualify to be polled.”

In that (Dec 2005) poll, Kalonzo Musyoka was in first place with a 35% vote. We should remember that Kalonzo had been selling himself a presidential material while Raila preferred to be considered for Premiership. It is only recently that Raila declared his interest in presidency. That Kalonzo beats Raila –the most vocal LDP chief hawk- should not be shocking. It is possible for somebody like Balala to beat Otieno Kajwang’ in a popularity contest.

In my opinion, Kalonzo is a more popular presidential candidate as compared to Raila because of the image Kalonzo has created. Kalonzo presents himself as a cool, diplomatic, intelligent, honest, and loyal guy. This is in contrast to Raila’s self created image of a forceful, cunning, coup loving, violent, and arrogant guy. When you complement this with what others think about him the result is a disaster.

A glimpse of past Steadman polls

Referendum
Steadman’s prediction
No – 42
Yes – 32
http://www.eastandard.net/archives/cl/hm_news/news.php?articleid=30953&date=22/10/2005

Results from real polls
No – 57
Yes - 43
http://www.eastandard.net/archives/cl/hm_news/news.php?articleid=32609
Note the difference in both results?

December 20006
Kibaki – 42%
Kalonzo – 20%
Raila – 14%
Uhuru – 5%
Ruto – 3%
http://www.eastandard.net/hm_news/news.php?articleid=1143962882

July 2006
Kibaki – 30%
Kalonzo – 27%
Raila – 14%
Uhuru – 13%
Ruto – 5%

http://www.eastandard.net/archives/cl/hm_news/news.php?articleid=1143955691&date=22/07/2006

March 2006
Kalonzo – 34%
Kibaki – 24%
Uhuru – 22%
Raila – 11%
Ruto – 4%
http://www.eastandard.net/archives/cl/hm_news/news.php?articleid=38826&date=1/4/2006

December 2005
Kalonzo – 35%
Kibaki – 26%
Uhuru – 17%
http://www.eastandard.net/archives/cl/hm_news/news.php?articleid=33956&date=21/12/2005

June 2005
Presidential elections
Kalonzo – 29%
Uhuru – 29%
Kibaki – 15%

Premiership
Raila - 35%
Kalonzo – 11%
Ngilu – 11%
http://www.eastandard.net/archives/cl/hm_news/news.php?articleid=22565&date=11/06/2005


July 2004
Uhuru – 27%
Kibaki – 24%
Kalonzo – 18%
Raila – 4%
http://www.eastandard.net/archives/july/sat10072004/headlines/news09070420.htm

2 comments:

Anonymous said...

The elephant in the room is that even with a large mass of MPs around him, even with a real campaign against Kalonzo, even with a tribe so much larger than Kalonzo's and with a war chest and national fame much greater than Kalonzo's Raila still cannot beat Kalonzo in an opinion poll.

Even when Kalonzo had NOT declared his candidature, still Kalonzo remained more popular than Raila. This is the question ODM-K are not addressing, and as our poll shows at KenyaImagine, Kalonzo may be the decisive factor at the election.

emmo

patriot said...

Steadman Cooking up Numbers.

A casual addition of President Kibaki’s support per province as released on Friday then divided by eight (the number of provinces) showed that Kibaki had scored 39.5 per cent, Raila had 47.6 while Kalonzo had 8.5 per cent.

http://eastandard.net/news/?id=1143977974&cid=159

Presidential voting by province
Province Kibaki Raila kalonzo
Nairobi 46.3 39.3 12.1
Central 92.8 5.0 2.0
Coast 34.7 51.5 9.7
Eastern 48.9 6.1 43.8
Nyanza 11.9 85.3 1.8
Rift Valley 39.3 55.8 3.9
Western 20.7 72.8 1.4
North Eastern32.2 64.4 3.4
Total 43.2 43.6 11.4

Kibaki average: 46.3+92.8+34.7+48.9+11.9+39.3+20.7+32.2=326.8/8 = 40.85

Raila averages: 39.3+5.0+51.5+6.1+85.3+55.8+72.8+64.4=380.0/8 =47.525


The weighting has already been taken care of by using a representative sample size based on registered voters per province. The larger the registered population in a province the larger the sample size!

That means that the results can be deduced by simple addition and division in two ways:

1. Because the sample sizes have been weighted you can add the total number of votes of each presidential candidate and divide by the total votes to get the percentage of the candidate!
2. Because the sampling has been weighted you can add the percentage provincial totals of each candidate and divide by the number of provinces.