Fast forward to the next election, and it was clocks all about, 'Masaa ni ya Ngilu', they said and all the signs bore them out. The Kitui Central MP drew massive crowds across the country, she got massive air time in the media and her message of social justice resonated with the poor and the marginalised across the country. In addition, she was a woman, and one representing an ethnic constituency that had never before challenged for State House. The media praised her and we were wowed, she was predicted to show a very powerful performance especially as she was backed by Anyang' Nyong'o, MP for Kisumu Rural and the pro-poor Social Democratic Party.

In the end neither of these candidates won the big prize. Matiba ended up in second place despite respectable showings in the Rift Valley, Central Province, the Coast and Western Province. Ngilu pulled up in fourth place, with the support of those who had showed themselves eager in their support vanishing into thin air, and leaving the oracles with egg on their face.

So it is that Kenyans must take the published opinion polls with a pinch of salt. I am not in any way suggesting as our politicians are likely to do that the polling companies are putting out doctored results. There is without doubt every likelihood that the wide variance in the pollsters' results is in part informed by their political leanings or the associations of their owners, but let us analyse the polls themselves. 

One reason does stand out, however, for the inconsistencies in the predictions. A central statistical problem with opinion polls is selection bias. Regular opinion polls, still sample only a small fraction of voters. The average number has been around 2,000 voters spread around the country on the basis of either population or more specifically by number of registered voters. With a completely homogeneous population, even a small sample is sufficient. Sample size depends directly upon the variability of the population: the more heterogeneous a population, the larger the sample required. In a heterogeneous population, like Nairobi's or Mombasa's or even in the border districts like Kajiado, Nakuru, Molo, Trans-Nzoia and Uasin Gishu, poor selection of the sample can tilt results in either direction. It is impossible to ensure that a sampling of Nairobi as currently undertaken takes into account the different ethnicities, tribal biases, the different social classes or indeed whether or not the respondent is likely to vote in Nairobi or upcountry. These are massive influencing factors that an ordinary sampling cannot take into account. What this means is that under the current arrangement, some constituencies are likely being over-sampled and others under-sampled. The solution would be to increase the overall sample size so as to better capture variations in these especially mixed areas.

Another factor likely to throw the results of opinion polls is the likelihood of participation in opinion polls. With many Kenyans not owning telephones, opinion polls in our country are conducted mainly face-to-face; either in the street or in people's homes. Certain constituencies are more eager to participate in such polls, not just because they have more time at their disposal but also because they are more passionate about politics than others, they feel a greater need to document their approval or disapproval. Such constituencies are also more predisposed to sabotaging the polls by providing a false vote, for example claiming to be a registered voter when one is not. These non-response and response biases can cause large shifts from straw-polling results come election day  and are in my opinion  truly unavoidable.  Given the fact that in a country like ours certain communities, and certain population segments are more likely to be politically passionate than others, there is no accounting for what the actual voter predilections are.

It is also important in analysing the opinion polls that we factor in the effect of social pressure on opinion poll results, and the very different atmosphere of the voting booth where each voter will vote unimpeded. This is especially important given the euphoria surrounding the Orange Democratic Movement. In many parts of the country, even those who feel it would be too great a risk to bet on Raila Odinga, or who perceive the Mwai Kibaki government's policies as having been beneficial to them may feel pressured to back the seemingly popular candidate, even as reservations are held to be exercised on Election Day. The opposite is also true in many areas where voters may hold specific grudges against Mwai Kibaki and a preference for the ODM candidate. Again, such a bias may only be confirmed in the lonely confines of the election booth. This spiral of silence, conformity to the more passionately held opinion makes for a very different picture in polling than what the results will show on Election Day. Steadman pollster Tom Wolf alluded to this in a television interview recently where he said that the fact that the ODM head Raila Odinga's main backers were young Kenyans who may not necessarily turn out on election day, while President Kibaki's supporters predominate in the 40 and over age segment which has proved to be a reliable voting bloc.

Another result of this pressure to conform is the peculiarly small number of undecided voters. It is unlikely that so few Kenyans have made up their minds, or that these positions are held as stubbornly as the pollsters would like us to believe. The constant changes

In the final push before Election Day, we will also see the effect of already published polls on final voting patterns. First is the fact that the polls illustrate the futility of a vote for Kalonzo Musyoka and the ODM-K. As Election Day approaches, it is likely that many of Kalonzo's supporters, seeing that their votes may decide the election, will switch their votes to their second preference- the lesser of the two devils in common parlance. This is especially true because of the manner in which Kalonzo Musyoka came to be a presidential candidate and the demeaning way that he was treated by the ODM candidate and his supporters. It is also instructive that the Mwai Kibaki team has many Ukambani politicians in it unlike the Raila Odinga team which only possesses the broken Charity Ngilu. How this plays out remains to be seen, but Kalonzo's 10-15% of votes represents a great hunting ground for either of the two leading candidates as is seen by the president's recent forays and reception in Ukambani.

Plus, if the polls continue to show such a slim lead for the ODM, the PNU base is likely to come out in large numbers believing that the result can be overturned. Conversely, the ODM crowd may get demoralised having had their lead squandered by a few crucial injudicious decisions from their leaders or if the last opinion polls are anything to go by, be carried away by their impending triumph and as a result lose the race.

All in all, the suggestion by the pollsters that the positions taken by the voters are rigid is unlikely to hold true. The 2005 referendum cannot be used as a test case. The ramifications of either vote in that time are not the same as they are now, and the government side campaigned so lethargically that many thought it had decided to lose the referendum. Also vital to take note of, is the fact that in 2005 the charismatic churches and Kenyan Muslims were united in their rejection of the proposed constitution. This time, these two groups will likely end up on different sides of the political divide especially following the shadowy MoU between Raila and NAMLEF. In 2005, President Moi and his treasure were waged against the government, this time they are aranged in support of the government. The Akamba vote was firmly in the ODM's pocket, this time it has been taken away from the party. The economic competence of the ODM was not on the stand in 2005, it is today and neither did the sceptre of Majimbo hang over us.

The voters have not yet spoken, and it is wrong that any organisation pretend that they have. There is still some time before Election Day, and a lot of ground may still to be covered.