Why opinion pollsters interview only a few people

By MUNGAI KIHANYA

The Sunday Nation

Nairobi,

10 December 2006

What is the average height of a Kenyan adult male? To find the answer, most people would just go around the country with a tape-measure and take the measurement of all men. Then these values would be added up and divided by the total number of people.

This exercise is impractical though it might produce a very accurate result. But is it necessary to take the measurement of each and every adult male Kenyan?

Suppose the height is measured accurate to the nearest centimetre. The sum of the measurements from the first ten men might come to, say, 1,732cm. Diving this by 10 gives an average of 173.2cm. However, since the data was only accurate to the nearest centimetre, it is only reasonable to round off the average to simply 173cm.

Now suppose we measure the eleventh man and find that he is 180cm tall. We add this to the previous total (1732cm) to get 1912cm. Divide the new sum by eleven and the result is 173.8cm – which, again, we round off to 174cm.

True; the new average is one centimetre greater than the older one. But is that difference significant? A change from 173 to 174 is equivalent to a 0.6 percent increment. I leave the decision to you, the reader.

But clearly, it is unlikely that the twelfth or twentieth or hundredth or thousandth or millionth man will make a difference greater than one percent to the average determined from the first ten. Thus it would not be necessary to go to each and every adult male.

However, it is said that Maasais are generally taller than Kikuyus and that Kambas have the shortest men in this country. Therefore, it would be advisable to include a few people from each community in the exercise. Thus we might go to each district and measure the heights of only ten men each. The total number would be about 700 and it would produce a very good result.

This is the principal applied by the organisations that conduct opinion polls. They recognise that the average results would not be significantly different if they interviewed 20 million or 2 million or 200,000 or 20,000 or 2,000 people. Thus naturally, they go for the smallest number.

Furthermore, the pollsters also recognise that, like heights, political opinions vary from one community to the next. Therefore, a few people are interviewed in each region of the country.

 Now, if the opinion of a few thousand people is representative of the average view of the whole population, does it mean that it is not important out go out and vote? NO! The results of an election are not based on the average outcome but on the total vote count.

Think of it this way; suppose you measured the height of 10 Maasais and 20 Kikuyus. The sum for Kikuyus would be greater than that for Maasais. Would you then conclude that Kikuyus are taller than Maasais?