Household and Income Expenditure Survey 1993-1994

There are essentially two sampling procedures that were followed in the HIES 1993/94. One for Walvis Bay and one for the rest of Namibia. These will be addressed in turn.

The HIES 1993/94 for most of Namibia follows a two stage sample design, taking random draws of geographical areas before randomly selecting households within that area. To do this, the Central Statistics Office (CSO), under the National Planning Commission (NPC), had to first develop a master sample frame.

To develop this master sample frame, a set of geographical areas, Primary Sampling Units (PSUs), was created and contained, on average, between 80 and 200 households. These areas were built from the Enumeration Areas (EAs) prepared for the 1991 Population and Housing Census. Small EAs were combined with adjacent EAs to form PSUs of sufficient size. The rule applied was that the number of households in a PSU according to the 1991 Population and Housing Census should be at least 80 households. About 1300 of the 1695 PSUs are made up of single EAs from the 1991 Population and Housing Census while about 400 PSUs are formed by joining two or more EAs. The 1695 PSUs, covering the whole of Namibia (except the Walvis Bay area), were separated into strata of PSUs by region and by rural, small urban and urban areas. The stratification into rural, small urban and urban areas was based on a classification of enumeration areas conducted during the preparations of the 1991 Population and Housing Census. (Note: A different definition of rural and urban areas is used in the statistical reporting from the 1991 Population and Housing Census and the HIES.) The urban areas in the Khomas region and some urban areas in the Otjozondjupa region were further stratified into high income and middle/low income areas. In this way 32 strata were created for the sampling of PSUs As a result of the way the PSUs were created, the number of PSUs of the master sample frame in each region and in each stratum is roughly proportional to the number of households in the region and in the stratum respectively as estimated in the 1991 Population and Housing Census.

Having developed the master sample frame, the HIES 1993/94 was sampled according to a Probability Proporitional to Size (PPS) of PSUs method in the first stage and a fixed size equal probability sample of households in each selected PSU in the seond. 192 PSUs were selected in the first stage as the master sample.

Initially the master sample was proportionally allocated over the strata in the master sample frame according to the number of households in the 1991 Population and Housing Census. However, some modifications of the allocation were made based on the following:

• The variation between households in income level seems to be generally larger in the urban areas than in the rural areas.
• The survey costs are considerably lower in the urban areas.
• There should be at least 10 PSUs sampled from each region to allow for reasonably good statistics from each region.

It was deemed necessary to have a slight oversampling in urban areas and in one region (Omaheke). A proportional allocation of the 192 PSUs over urban/rural areas gave 66 urban and 126 rural PSUs. But, given the above specifications, the selection of the master sample resulted into 81 urban and 111 rural PSUs.

In the second stage, households were selected from the chosen PSUs. A list of households for each PSU was prepared during a separate listing exercise. The listing was carried out as closely as possible to the start of the data collection in a certain PSU i.e. the month before the HIES survey month of the PSU. The list of households in the PSU was used as the sampling frame for the selection of households and a systematic equal probability random sample of 24 households from each PSU was drawn.

As initially mentioned, sampling was done slightly differently in Walvis Bay. This is because the region was not integrated in Namibia until 1 March 1994 and could therefore not be included in the HIES before that date. For planning and logistic reasons Walvis Bay was included in the survey somewhat later - from May 1994. This means that Walvis Bay was included in the HIES during the last six months of the survey year. The sampling procedure was different from the rest of Namibia. In Walvis Bay the municipality authorities have for administrative purposes created computerized registers of the households in all the three main town areas - Central Walvis Bay (incl. Langstrand), Kuisebmund and Narraville. Most of the households in Walvis Bay are covered by these registers. There are some areas, however, which are not covered by the administrative registers. These areas are the hostel areas of Walvis Bay and the area along the Kuiseb river where the Topnaar population lives. To cover also these population groups, CSO conducted listing of the households in these areas. For security reasons all the hostel areas could not be listed but some areas had to be excluded from the HIES. The number of listed households in the hostel areas was 99 and probably about the same number of households was not listed. The number of Topnaar households listed was 73.

Altogether 144 households from Walvis Bay were selected by mainly a stratified one-stage sample design to be included in the HIES sample. 24 households were sampled during each of the six months (6 * 24 = 144). Separate strata were defined for Central Walvis Bay (incl. Langstrand), Kuisebmund (excl. the hostel areas), Narraville, the hostel areas and the Topnaar population and altogether 36, 54, 36, 6 and 12 households respectively were sampled from each stratum. (During May-July 1994 10 households were selected each month in Kuisebmund excluding the hostel areas. During August-October 1994 only 8 households were selected in Kuisebmund excluding the hostel areas while 2 households were selected in the hostel areas.)