Poor Methodology Blights Study on DA and Apartheid

Article excerpt

BYLINE: Herman W Kruijsse

Several newspapers cited a recent survey that asked respondents between the age of 15 and 34 and from different ethnicity and gender whether they believe that the DA, if elected, would bring back apartheid. Apparently 52% of blacks, 26% of Indians, 21% of coloureds and 19% of whites believe the DA would bring back apartheid.

The researchers of the above survey state that the results are nationally representative in terms of age, gender and race.

The results are based on the responses of 3 009 respondents. There are four main factors: age, race, gender and provinces. The province factor is necessary to control the variety within and between provinces. For example, the population of Limpopo consists almost 97% of black people whereas black people in Gauteng, whose population is almost triple that of Limpopo, comprise 74% of the population.

"Provinces" has nine categories, "gender" two categories and "race" four categories. "Age" runs between 15 and 34 years old. Since it is likely that age is observed as belonging to an age class, I must assume that at least three age categories were used: 16 to 20; 21 to 27 and 28 to 33.

Therefore, if every category is to be observed, at least 12 males and 12 females per province are required to have a single opinion per province. Thus we would need a "national" sample size of 216 (2x3x4x9) to have one valid observation per province.

It is evident that a single black or white or coloured or Indian female or male respondent of a particular age category is not representative for the opinion of people with similar demographics in a province, neither for that matter would nine respondents be representative of the whole country.

Would you think that 14 per province would be representative? For that is what the sample size of 3 009 used in the survey implies.

Interviews were carried out on cellphones. This raises several questions. Did a random generator generate cellphone numbers that were called by trial and error and were people asked to state their demographics prior to an interview? Was the cellphone number of the survey group advertised so that people could respond? Were numbers obtained by some databases of which owners of cellphone numbers agreed to have their numbers used for such purposes?

We do know that there is no public directory of cellphone numbers; they are private.

Either way - legal, illegal or unethical - interviewing by cellphone is prone to non-random sampling. …