Living Conditions Monitoring Survey II 1998

Sampling Frame and Stratification
The country is made up of 9 provinces comprising 72 districts delineated by the Local Government Administration. Previously there were 57 districts in Zambia. 15 new districts have been created. Central Statistical Office has delineated these districts into Census Supervisory Areas (CSAs) and then these into Standard Enumeration Areas (SEAs) for the purposes of conducting censuses and sampling for surveys. Each CSA is made up about 3 SEAs. The list of CSAs and SEAs by province & district constitute the sampling frame for CSO censuses and surveys. The sampling frame comprises 4,193 CSAs of which 3,231 are rural and 962 are urban and 12,999 SEAs. The frame of CSAs and SEAs is arranged by province, then by district within a province,then by rural/urban within a district, then by centrality within rural/urban, and finally by low, medium or high cost for urban SEAs. The frame also contains information on the number of households and the population size per SEA and this is what was used when selecting the sample using the probability proportional to size (PPS) method. The number of households and the population in the frame is based on the 1990 population census. To boost the data from the survey to 1998 population parameters the weights calculated were multiplied by a factor equal to the estimated population growth from 1990 to 1998. This was done at the district level.

The classification of centrality is shown below:-
Centrality Classification:-

1. Areas within Lusaka city.
2. Areas within Ndola city.
3. Areas within Kitwe city.
4. Areas within 50 Kms radius outside Lusaka, or Ndola, or Kitwe cities.
5. Areas within provincial capitals.
6. Areas along Southern to Copperbelt line of Rail (within 30 Kms radius).
7. Areas along Northern line of Rail (within 30 Kms radius).
8. Areas within 30kms radius outside provincial capitals.
9. Areas within district centres.
   10.Areas within 30 Kms radius outside district centres
   11.Remote areas.

Areas within cities, provincial capitals and district centres is equivalent to the urban part of the town.Within the rural SEAs households have been classified on the basis of the scale of agricultural activities into small scale, medium scale, large scale, and non-agricultural households.The urban SEAs have been classified into low cost, medium cost or high cost depending on the type of housing in the area.The local government administration has classified localities into low, medium and high cost based on the required housing standard. The urban SEAs were classified into low, medium and high cost areas based on a combination of the local government and CSO criteria. All urban SEAs were physically visited by CSO mapping staff with locality classification from local government and determined whether the SEA was low, medium or high cost based on the local government definition and the actual observation of the mapper. The mappers were trained on how to make this determination. Households within rural SEAs were classified into small scale, medium scale, large scale, and non agricultural households after the listing operation.

Sample Size: Out of a total of 12,999 SEAs in the frame, a sample of 820 SEAs were selected for the Living Conditions Monitoring Survey (1998) representing about 6% of the total. The urban stratum was allocated 328 SEAs and the rural stratum was allocated 492 SEAs. The total number of households enumerated were 8520 in rural areas and 8220 in the urban areas.The total number of persons who fell in the sample were 45989 in rural areas and 47480 in urban areas.All the 72 districts in Zambia were covered in the survey on a sample basis.

Sample Allocation: Sample allocation was done using the "Probability Proportional to size" (PPS) method. This entailed allocating the total sample (820) proportionately to each province according to its population share.Thereafter, allocation of the provincial sample was done proportionately to each district according to the population share from the provincial population. Similarly allocation was done by centrality within a district. For example, Mkushi district was allocated 10 SEAs by the PPS method. The district has four centrality classifications (9, 7, 10, and11). The number of SEAs under each centrality classification in the frame were summed up. The next step was to determine the share of each centrality group of SEAs from the total number of SEAs in the frame under Mkushi district. The corresponding proportions were used to allocate the sample to each centrality category. However, the final allocation was plus or minus depending on what was obtaining in the frame. For example if 1 SEA was to be allocated to centrality 9 (District centre) by using PPS and yet there is low, medium & high cost SEAs under centrality 9 in that district, the number of SEAs selected was 3 (one from low, and the other two from the medium & high cost SEAs). Not all centrality classifications obtain in all districts, for example, Lusaka district had all the SEAs fall under centrality 1 (Lusaka city) in the frame. Therefore the entire number of SEAs allocated to Lusaka district was selected from this category. The minimum size for each district sample was 7 SEAs, meaning that even the smallest district was allocated at least 7 SEAs.

Sample Selection: Sample selection was done in two stages. In the first stage, a sample of SEAs was selected within each stratum (centrality) according to the number allocated to that stratum. The second stage comprised selection of households from each sample SEA according to the number of households recommended after a complete listing of all households in the sample SEAs. Thus SEAs formed primary sampling units. The unit of analysis was the household.

Selection of SEAS: After sample allocation was done, selection of the sample SEAs from the frame followed. The allocated number of SEAs were selected at centrality level using the PPS method.

Selection of Households:In each selected SEA, households were listed and each household given a unique sampling serial number. A circular systematic sample of households was then selected. Vacant residential housing units and noncontact households were not assigned sampling serial numbers. Selection of sample households was done by supervisors in the field and they were required to select the following numbers of households:
30 households from SEAs with sample Micro-projects (whether rural or urban).
25 households from urban SEAs (without sample micro-projects)
15 households from rural SEAs (without sample micro-projects). This number increased in rural SEAs where large scale farmers were identified.

In urban areas the required sample number of households were selected straight forwardly using the circular systematic sampling method. In the rural areas, 7 households were selected from the stratum of small scale farmers, 5 from medium scale farmers, 3 from non-agricultural households, and all large scale farmers if any were found in the SEA. Therefore, the number of selected households from a rural SEA was more than 15 where there were large scale farmers. In Micro-project areas the number of households to select was double, 14 in the small scale category, 10 in the medium, 6 in the non-agric, and all large scale farmers.

The circular systematic sample selection method was used in both rural and urban SEAs.The circular systematic sampling method assumes that the households are arranged in a circle and the following relationship applies (Kalton G., 1983):-
Let N=nk
were, N=Total number of households assigned sampling serial numbers in a stratum.
n=Total sample required from a stratum.
K=The sampling interval in a given stratum calculated as K=N/n.

Therefore, for urban strata K=N/25 or K=N/30 (for micro-projects areas).

And for rural strata, K=N/7 in the small scale stratum, K=N/5 in the medium scale stratum, and K=N/3 in the non-agric stratum. K was not calculated in the large scale stratum as all large scale households identified in an SEA were enumerated. In most cases there were no large scale farmers in an SEA and in a few cases there was no more than one. In the case of micro-projects areas the corresponding K'S were N/14, N/10 and N/6 respectively. The N in the rural SEAs differed from stratum to stratum within an SEA.

The steps in selecting households were as follows:-

(i) A random number was obtained using a table of random numbers. This number was between 1 and N (both inclusive). The urban SEAs had one random start per SEA while the rural SEAs had three random starts per SEA (one for each stratum except the large scale stratum).
(ii) The sampling interval (K=N/n) was calculated for each urban SEA and for each stratum in the rural SEAs.
(iii) The sample number of households required was then selected using the circular systematic method. The household whose sampling serial number corresponded to the random start was the first to be selected. Then K, the sampling interval was added to the sampling serial number of each selected household in the respective strata until the required n (sample size) was achieved. All in all 8220 urban households and 8520 rural households were selected.