Opinion polls don’t add the size of ugali in my plate!
The Sunday Nation
Nairobi,
27 January 2013
I do not know who will win the coming presidential elections! I say that because many readers have asked me to show them how they can predict the winner. Saying that they do not trust the opinion polls, the said readers would also like to know whether I think the results are genuine.
Well, I don’t know whether the results are genuine because, truly, I don’t care! I am not a candidate in the elections and I do not work for any campaign secretariat. Therefore, the poll results do not affect me in any way. As former President Moi would say: “They do not add the amount of ugali in my plate!”
For that reason, I do not spend any time or effort trying to figure them out. I do, however, analyse each candidate with the aim of finding out whom I most prefer as my president – their predicted chances of winning do not enter into that analysis at all.
Nevertheless, the biggest problem that people have with opinion polls is
that the samples used appear to be too small – typically 1,500 to 2,500
in
The answer is a resounding YES; but only if the sample is adequately random. This means that the polling process must give each member of the target population (e.g., in the case of political opinions, all registered voters) an equal chance of being questioned.
To understand and appreciate how this is so; suppose you poured a little sand (say, one cup) in a sufuria full of water. Obviously, the sand will settle to the bottom of the vessel. If you now wanted to know the percentage of sand in the water, you would stir the mixture thoroughly and then scoop out a small quantity – even a spoonful will be enough! Stirring gives every grain of sand an equal chance of being picked in the sample.
Then you would weigh this sample, let it settle, pour away the water and weigh the sand alone. The weight ratio you will get will be very closely representative of the situation in the (much) larger sufuria.
The same principle applies to people’s opinions. The pollster must adequately “stir” the population in order to get a representative sample.
But there still remains the question of whether poll results affect the outcome of the election. I think they do. My view is based on the number of times I hear people saying that they do not want to “waste” their vote by giving it to a candidate who is unlikely to win.
For that reason, I also hold the position that opinion poll results should never be made public. Apart from undermining the democratic principle of free choice, I think that publicising them also dents the reputation of the pollsters.
Perhaps it is time the leading research companies did a poll with the question: “Do you believe in opinion poll results?” If the majority of respondents say “no”, then the researchers should think carefully about the value they add to our society when they publish their results.
 |
 |