Gordon Muller has been calling for an overhaul of AMPS, particularly citing it is impossible to convince clients visiting Maponya Mall that AMPS is correct in reflecting only two LSM 10 respondents in Soweto. SAARF management responds.
The SAARF LSM has been a vital tool in the industry’s marketing segmentation toolbox for over 20 years. Based on multivariate statistics and using a basket of variables that are strong discriminators from the AMPS survey, it has proven to be a very useful and sensitive device, not only for segmenting a market, but for tracking the development of a changing nation.
If it’s so sensitive though, why are we not picking up more LSM 10 black people in Soweto? That’s what some people in the industry have been asking. Why does AMPS, which carries the LSM variables, only pick up and interview a couple of LSM 10 blacks in Soweto? Surely the sample should be bigger so that we pick up more of this market?
To answer these questions, you need first to understand how research of the magnitude of a national survey is carried out. To begin with, research does not produce facts, but rather gives estimates of the most probable scenario. If you want absolute fact on a national basis, you’d need to interview the entire population, which even the Census, at a cost of around R2 billion, cannot do.
Instead, a sample stands in for the entire population. SAARF interviews 25 000-plus people in 25 000-plus households every year. It then grosses up these responses to the greater population. To be able to do this correctly, a multi-stage area stratified probability sample is used.
This type of sample will give you results that most accurately represent all the parameters of a population, with the right proportions in each geographic area. Unlike quota sampling where, for example, you decide you want to research LSM 10s and then actively seek them out to interview, random stratified probability sampling uses address registers or GPS co-ordinates to randomly choose respondents across the country – from metropolitan areas right through to the deep rural areas. And unlike quota sampling, probability sampling allows you to gross up your results to the entire population.
“We don’t go out specifically to find a certain number of LSM 10 people in Soweto,” says Dr Paul Haupt, CEO of SAARF. “If they are there, they will be picked up in relation to their natural occurrence. We don’t make assumptions based on what we think should be happening in an area, and then fiddle with the sample to ensure we get the answers we’re looking for. After all, if we knew the answer, why would we be doing the research?”
What then is really happening in the top end of the market in Soweto?
LSM 10 is not a large group, accounting for just 6.4% of the total South African population. It makes sense that out of a total 25 160 interviews in the latest AMPS release (AMPS June 2011), only 2 959 (or 11.8% of the sample) turned out to be with LSM 10 respondents. By research standards, however, this is in fact a very large, robust sample. Consider that such rigorous surveys as futurefact and Finscope are conducted off total samples of a size similar to just the LSM 10 sample in AMPS.
When working with what is already a select group of individuals such as the people in LSM 10, you are liable to run into problems if you over-segment that group, for example, by analysing a further demographic sub-group within a single area such as Soweto. Hence, if you want to look at black LSM 10s in Soweto only, you’re looking at less than one percent of all black LSM 10s. This is a classic case of over-segmentation.
“Because the total AMPS Soweto sample is quite large, at more than 500 people, we accurately know that the percentage of black LSM10 people living in Soweto is small,” says Haupt. “Do we therefore boost the sample in Soweto, specifically into the LSM 10 range, at considerable cost, to specifically focus on a market of less than 1% of black LSM 10s?
“I suggest that those wanting to target Joburg’s black LSM 10s – and for most brands, this would be over-segmenting – should rather fish where the fish are, and not only look in Soweto,” says Haupt. Because South Africa has changed over the past 15-odd years, and black LSM 10s are not necessarily living where you might think they ought to.
Today, 59% of black LSM 10s live in metropolitan areas, and 15% live in greater Johannesburg (Alexandra, Johannesburg, Sandton, Randburg and Soweto). And of those who live in Johannesburg, 38% live in Sandton, while a further 24% live in the northern suburbs.
In the introduction to Jodi Bieber’s book ‘Soweto’, published in 2010, Niq Mhlongo says, talking about the Panyaza Butchery Chisa-Nyama in Soweto’s White City neighbourhood: “The coolest of the coolest in the trendiest of cars hang out here. Many of these people moved to the suburbs but return every weekend as they prefer the vibe.”
Maybe this is then a case of AMPS and the LSMs actually doing what they’re meant to – reflecting the reality of our society instead of confirming our preconceptions.
The changing face of South Africa
The SAARF Living Standards Measure (LSM) is a sensitive barometer of societal change and development. Since democracy, the wealth status of the South African population has changed considerably for the better.
In the past 10 years alone, there has been a steady movement into the middle classes out of the lower LSMs. Between 2001 and 2011, the proportion of the population in LSM 1-4 fell by 48%, while LSM 5-7 grew by 57%. LSM 8-10 also grew by 45%.
The most marked changes have happened in the black community, where LSM 5-7 grew by 82.5%, and LSM 8-10 by 344% over the 10 year period.
The wealth status of provinces has also changed significantly, when comparing AMPS 2001 and AMPS June 11 (July 2010-June 2011):
• The Western Cape and Gauteng show the biggest decline in
LSM 1-4s, down 78.4% and 82.9% respectively.
• Large gains in LSM 5-7 were seen in Limpopo (326.7% up), North West (131.0%), the Free State (98.7%) and Mpumalanga (97.1%).
• Limpopo’s LSM 8-10 population rose by 87.5%, while North West (75.3%), KwaZulu-Natal (69.6%) and the Eastern Cape (64.3%) were also much stronger in the top LSMs.