To get 13,795 people in Gauteng to answer 200 questions about their lives, the Gauteng City-Region Observatory had to knock on just over 28,000 doors. That is a hit rate of under 50%. Each successful interview runs about 45 minutes. The whole exercise takes years to plan, run and clean.
That is the unglamorous foundation of one of the richest pictures any province in South Africa has of itself. And, as Christian Hamann explained at The Outlier’s Out to Lunch session last Wednesday, it is getting harder to maintain.
Hamann leads the quality of life studies theme at the GCRO, a research body set up in 2009 as a partnership between the Gauteng provincial government, Wits, UJ and organised local government. The GCRO sits in the gap between academia and policy, producing work nuanced enough for journals and digestible enough for officials who do not have time to read journals. Its flagship product is the Quality of Life Survey, now in its seventh iteration.
The slow business of asking people things
The survey is not a poll. It is built on a two-stage spatial sample that covers all 529 wards in Gauteng, every nook and cranny of the province in Hamann’s phrase. Within each ward, a handful of enumeration areas are randomly selected. Within each enumeration area, individual dwellings are randomly drawn from a dataset of plotted residential buildings. A trained fieldworker has to physically arrive at the address, work out how many households are there, randomly select one, randomly select an adult inside it, and conduct the interview on a tablet.
The result is a dataset that runs from the visible to the felt. Dwelling type, services, transport, employment, healthcare on one side. Satisfaction with water supply, opinions on government, sense of where the country is heading on the other.
The questionnaire is also stretched and pruned over time. About 200 questions are retained for comparability across iterations. A small portion is held open for new research interests or whatever the moment demands. The COVID round added questions about lockdown experience and job loss. The next round added vaccine hesitancy. An earlier round, at the height of load-shedding, asked households whether they were considering solar.
The case for face-to-face
The GCRO has stuck to in-person interviews through all seven iterations, but the pressure to switch is real. Phone and web surveys are cheaper. Gated estates and security complexes make physical access a problem in parts of the province.
Hamann is wary. The literature still favours face-to-face for data quality, partly because a fieldworker can rephrase a confusing question. He gave a small example from the GCRO’s behind-the-glass testing, where researchers watch volunteer interviews through one-way glass. The team had been asking respondents “What is your age?” and assumed it was simple. One respondent looked blank. Reworded as “How old are you?”, the question worked. Same data point, different cognitive load.
Switching survey modes also breaks the longitudinal value. Sixteen years of comparable answers is not something you replicate cheaply. The likely middle path, Hamann suggested, is keeping the spatial sampling but finding ways to confirm a respondent’s willingness before sending a fieldworker to their door.
Visuals that get out of the way
Good data visualisation, Hamann argued, is mostly not about software. It is about empathy and translation. He showed an example from a 2007 analysis of household income growth between the 2001 and 2011 censuses. The output was a map of growth rates per ward, accurate but useless to anyone without a statistical brain. The number 1.33 means nothing on its own.
The fix was to drop the numbers and recategorise wards: rich areas where income grew fast, poor areas where it grew slowly. Same data, same analysis, but suddenly the picture told a story about a widening gap. The discipline is in the framing, not the tool.
What the data is saying now
The forward picture from Quality of Life 7 is sobering. Of the dimensions that feed the GCRO’s quality of life index, four are in decline: safety, life satisfaction, basic services and government satisfaction. The drop in safety has been particularly sharp. Health and participation are roughly flat. Socio-economic status ticked up slightly after its COVID hit, but the recovery is uneven and fragile. People who got jobs back are often in less stable work, and one shock can reverse the gain.
Some of the GCRO’s clearest policy wins have come from this kind of fragility. COVID vulnerability maps, built quickly from existing data on overcrowding, public transport use and economic exposure, fed into early pandemic response. Earlier work on multi-dimensional poverty helped identify the 50 priority wards still referenced in provincial planning. A sharp decline in refuse collection in Lesedi pushed the premier of Gauteng to intervene.
What the GCRO cannot do is implement. It can show the picture.
The full Out to Lunch session with Christian Hamann is available to Outlier members in the recordings archive. Out to Lunch runs free webinars with researchers, journalists and analysts working with data. Subscribe to The Outlier to get invitations to the next one.