|
 |
 |
The Problem of Bias in Statistics | 14.10.2001 |
|
| |
 Are "Teletubbies" really that popular?!
When I was still working in England, I was given a newspaper article by a friend headed How children learn to hate the Germans. It was about a survey carried out by a psychologist from Aberdeen University, which stated that children in Britain become more and more prejudiced against Germany (and France, as well) from the age of ten onwards. If you are a German (well, European, actually) working in Britain, two thoughts spring to mind: firstly, how come the British in general are thick enough to tell their children that the war against the Huns still goes on? Secondly, if a report based on a poll says that 'children learn to hate the Germans', can I be sure that 'children learn to hate the Germans'? In other words, how reliable are public statistics?
All surveys are done by means of sampling; i. e. a small portion of the relevant section of the population is asked to allow estimates on the whole section. If the social group in question is rather large, as it is the case with British children between six and 16, of course the sample group has to be very representative of the whole in order to avoid bias. And that's just the problem. In my opinion, it all boils down to this: bias can come from everywhere and you will never get an out-and-out representative sample group.How can I know that this psychologist from Aberdeen chose his respondants randomly? For it is mentioned in the article that only white children have been interviewed and the number of coloured people in Britain is estimated at 4% of the total population. To avoid racial bias, one could pick every 20th name in a telephone directory and ask parents to send their children to the interview. But who is more likely to answer the phone, a well-off higher middle-class housewife or a Labour single parent who has to work all day to make ends meet? There again, bias comes into play. Recently I have read about a survey carried out by a major market research company to test the popularity of a special brand of baked beans with pork sausages.They chose an ordinary London district with normal income level and a mixed age-group as their research area. The only fact they overlooked was that the selected district, Golder's Green, has got a very high percentage of Jews. So it was hardly surprising that the number of people deciding in favour of the product was rather small because, understandably, most Jews are not too keen on pork sausages.
On top of that you have got the problem of bias caused by the insicerity of the interviewees. As part of a questionaire once a special number of men and women had been asked, how many members of the opposite sex they had already slept with.The average answer the women gave was 1.9. Concering the men it was 3.7. As in this survey the number of people asked was rather large, the numbers should have been roughly the same if the respondents had answered truthfully.
Therefore, when reading the paper, public statistics may be far more interesting than the Sports part, but they also should not be taken at face value. |
Author:
 Raphael |
|
|
 |
|
 |
|
|
|
