Department of Social Development Survey 2006-2008

The Baseline Survey
The 2006 baseline survey sought to conduct 400 interviews in each of the 14 ISRDP nodes and the 8 URP nodes. In order to allow for comparisons with the ISRDS (as it then was) baseline statistics published by Statistics South Africa in 2002, the 400 interviews for Maluti-a-Phofung were spread across the whole district municipality of Thabo Mofutsanyane. The adult population aged 18 and older according to the Census 2001 was used as the sample frame. For the ISRDP nodes, the sample was stratified by local municipalities to ensure sufficient interviews were conducted in each municipality. According to the principles of probability proportional to size sampling (PPS), a list of place names in each of the local municipalities was then generated as a starting point for the fieldwork. At each starting point in the ISRDP nodes five interviews were conducted. For the URP nodes, detailed maps at a ward level were generated from the Municipal Demarcation Board website. Again using the principles of probability proportional to size sampling (PPS), starting points across the different wards were identified on the maps. At each starting point in the URP nodes four interviews were conducted. At the end of the fieldwork phase a total of 8 387 interviews across the 22 nodes had been conducted. Once the information from each interview had been coded and captured on computer, the realised samples in each of the ISRDP nodes were weighted back to the actual population figures across each local municipality. It should be noted that on the one hand, 8 400 is a very large sample with a margin of sampling error of only 1.1%. However, when the data are analysed at nodal level, each of the 22 samples of 400 have a larger sampling error of 4.9%.

The Measurement Survey
The 2008 measurement survey sought to conduct 250 interviews in each of the 14 ISRDP nodes (except in Bushbuckridge and Maruleng, where 250 interviews were divided across the two nodes according to population size) and the 8 URP nodes. In order to allow for comparisons with the 2006 baseline survey, the 250 interviews for Maluti-a-Phofung were spread across the whole district municipality of Thabo Mofutsanyane. For comparative purposes, the sample frame (the adult population aged 18 and older according to the Census 2001) and list of starting points from the 2006 baseline survey was used. For the ISRDP nodes, the following steps were followed: The sample for each node was firstly stratified by local municipalities (to ensure sufficient interviews were conducted in each municipality). Within each municipality, the sample was then stratified by settlement type (rural versus urban). According to the principles of probability proportional to size sampling (PPS), a random list of place names in each municipality was then generated. At each place name, the fieldworkers were instructed to find a school (if multiple starting points at one place, subsequent starting points were at different schools or crèches). From the school, they then walked in the direction of dwellings and started at first dwelling - thereafter, every fifth dwelling was selected. The birthday rule was used to select the respondent at each selected dwelling - this random process seeks to interview the adult in the household whose birthday is next. For the ISRDP nodes, five interviews were conducted per starting point. For the URP nodes, the following steps were followed: ?. The sample for each node was firstly stratified by wards. Within each ward, the sample was then stratified by settlement type (formal versus informal types). Detailed maps at a ward level were generated from the Municipal Demarcation Board website. According to the principles of probability proportional to size sampling (PPS), a random series of starting points in each ward were then generated using a random grid of points. From the identified starting point, the fieldworkers proceeded in the direction of the centre of the node and interviewed at the first dwelling they came to - thereafter, every fifth dwelling was selected. The birthday rule was again used to select the respondent at each selected dwelling. For the URP nodes, four interviews were conducted per starting point. At the end of the fieldwork phase a total of 5 232 interviews across the 22 nodes had been conducted. Note, while 5 250 is a large sample with a margin of sampling error of only 1.4%, a nodal sample of 250 has a far larger sampling error of 6.2%.
For both surveys, sampling and weighting was undertaken by Ross Jennings of Strategy & Tactics.