Probably a dumb question but my googling skills failed me.

We’re having a census in SA soon and I’m wondering why it is necessary to try count everybody as opposed to an appropriately large sample size?
Is it a case of trying to count everybody to increase the probability of getting a decent enough sample size? Or is it just too difficult to subdivide the country into the appropriate areas to be sampled that is representative of density/ethnicity/wealth etc.

The main idea behind the census is to obtain data that are as accurate, comprehensive and reliable as possible. The problem with surveys is that they estimate whatever it is that they are designed to measure, and usually make certain assumptions that are not necessarily correct in order to simplify the survey design. Census data can be used to gauge the validity of survey results as well as to confirm assumptions about the continued homogeneity of previously-assigned statistical groupings, for example income groups in different living areas. Most importantly, census data form a much more trustworthy basis on which to determine current and future planning needs for the state. That’s why in some countries (e.g. UK) it’s mandatory to participate in census-taking.

Does that address your concerns?


And you’re saying that no sample is large enough to be good enough to be “accurate, comprehensive and reliable as possible”, we can’t statistically show that at some point we can apply a law of diminishing returns?

Finally I got a hit from google: http://www.sciencenews.org/sn_arc99/3_6_99/bob1.htm

The article is a bit old but some interesting quotes:

"Generally, the statistical community is very much in favor of trying to do sampling," says Donald B. Rubin of Harvard University.

Politicians don’t like this

Republicans vehemently oppose the use of sampling.

No surprises there.

Not that I assume that you know this personally, but does anyone know if it’s mandatory in SA?

You know you’re a geek when you find stuff like this interesting, for everybody else |-O

Statistics in this report were primarily taken from Census 1996 and Census 2001 and the more recent CS 2007. The comparative analyses of these sets of data should be regarded as an indication of broad trends in the area, as data integrity concerns regarding these data sets were raised. The South African Statistics Council was concerned about the following regarding the Community Survey (2007):

· Institutional population is merely and approximation to 2001 numbers and not new data;

· Unemployment in the CS (2007) is higher and less reliable because of questions that were asked differently;

· Grants data should be interpreted with great care;

· Income includes unreasonably high income for children – presumably misinterpretation of the question, listing parents’ income for the child;

· Distribution of households by province has very little congruence with the General Household Survey or last census.

The council states (Statistician General’s Response to the Star and other newspapers, 18 January 2008): “In the absence of a comprehensive sampling frame, it is difficult to determine whether the differences are due to sampling error, biases or the reality that has changed beyond our expectations. There may be other variables that will require similar warnings after further interrogation.”

A number of systematic errors were observed in the data, which include:

· An underestimate of men relative to women;

· An underestimate of children younger than 10 years;

· An excess of those aged 85+, in particular among men;

· Missing women aged 20–34 from the Coloured population;

· Misdistribution of the population by province;

· Excess of people aged 10–24 in Western Cape and Gauteng; and

· A shortfall of women aged 20–34 in Free State, KwaZulu-Natal and Limpopo.

Council found that the confidence intervals at some municipal levels were very wide. It then recommended that further analyses and investigation be conducted into the data to ensure that reliable data are released at district and municipal levels.

In a nutshell, the sample that is most representative is made up of the entire population. As a rule of thumb, in reliably estimating a single parameter for a large population, a sample size of around 2,000 is considered sufficient. However, a census measures lots of different parameters, many of which aren’t independent of one another, e.g. the value of your home and your household annual income. Census data also look at such interdependencies in order to build statistical models that are used for scenario modelling, contingency planning and strategic planning generally. Remember that a census is interested in counting the actual numbers and quantities involved, not just their relative frequencies. A survey won’t tell you how many people live in a given area although you could estimate it from survey results and other data, whereas a census will yield a much more trustworthy population count. It’s true that a random sampling of sufficient size and scope can give reliable results but thoroughness and historical legacy still hold sway. There’s also the problem that people need to be able to understand fully the questions they are being asked.

Participating in the census isn’t mandatory at present in SA.


I dont know for sure, but taking in account that they’re short of begging us to participate and not threatening us, I dont think its mandatory by law.

This is upon us in the next few days. I note with interest an ad on TV that they will be contacting non-conforming households via telephone, so, erm, where did they find my number? I scanned through the proposed questionaire and in all honesty, I’m not going to give any of the information requested, most of the personal details they can pull off the population register anyway, all of us have ID’s and birth certificates and marriages and deaths need to be registered as well.


what the hell does it matter if i have a tv? it should matter if i have a medical aid, running water, electricity. it just sounds like they are wanting a stick to beat the ‘advantaged whites’ with. the census said, that 99% of whites have tv’s, and only 2% of blacks, that means that whites are still oppressing the blacks, or whatever.
also, the radio ads say, you HAVE to let them into your home? good luck with that. i dont anybody in my home i dont know. i dont give a rat’s ass who you are. and good luck getting me at a time i’m even home. or getting hold of me through my fencing, big ass dogs… and what if i cant let you in? like, i have a huge dog, and my husband isnt home to handle the dog, should i let you in, and let my rottweiller rip your trousers off?
and isnt one of our constitutional rights, the right to privacy? cops need a warrant to enter your home, so why should some random census person in my home. why does he need to be inside? why cant i fill in his form outside?

You don’t have to let them into your home. They can drop off the form, you fill it in, then they come and collect it later.

I only want to know if it is required that I go to my hometown for this, because unlike some messiahs, mine is quite the distance away…

The community here would no doubt be interested to know that this year they aren’t asking for your religion, for the first time in an SA census. The race question is still being asked, though. I’ve said my bit about that here: http://dailymaverick.co.za/opinionista/2011-10-05-what-are-you

Yup, but since there’s no race classification board, your race is what you say it is. I can call myself black and according to the constitution no one is permitted to contradict me.

wonder who they will be sending into Orania…

Great, I was going to do my bit for atheist demographics. Now I can’t. :frowning:

If this is true, then it could instantly solve many companies’ BBBBBBBBBEEEEEEE requirements! Also, by handing in my gun and badge as a member of the “non-designated group”, I could vastly improve my career opportunities … interesting possibility! :slight_smile:


That’s what they do at the Free Market Foundation. Each year they classify all their staff as a different race. So this year, they’re all black, next year they’re all Indian and so on. No one can say them nay because it’s unconstitutional to do so. Other companies just don’t have the balls to stand up and do the same.

Um, yes - that’s basically the topic of the column.

I’m not going to participate.I’m a private citizen and wish to remain as such. My income, level of education, living conditions etc… are my private business and government doesn’t have any right to collect this information.