Household Budget Survey 2000-2001

Sample design: The sample of households interviewed in the 2000/2001 HBS was selected in two stages. In the first stage, 1,161 small areas called Primary Sampling Units (PSUs) were selected throughout the country. In the second stage, 24 households were initially selected in each PSU. The sampled households are located in the National Master Sample (NMS) of PSUs. The NMS is a generalised set of area units that can be used as PSUs for conducting various household surveys. It is a fixed sample of rural and urban clusters, which, among other things, make possible the performance of a continuous survey programme as well as ad hoc sample surveys.The NMS has four modules,A,A+B,A+B+C and A+B+C+D, which can provide urban and rural estimates at National, Zonal, Regional and District levels respectively. The HBS 2000/01 used Module A+B+C of the NMS comprising 621 urban EAs and 540 rural villages drawn from each of the 20 regions of Mainland Tanzania. In the second stage, 24 households were selected using systematic random sampling (SRS) from stratified lists of households compiled from each of the sampled PSUs. These lists were stratified into high, middle and low socioeconomic groups based on socio-economic data collected during the listing exercise. The stratification and selection of households was conducted in the NBS head office and interviewers were supplied with a list of pre-selected households for interview.

Sample frame and sample selection: (a) Rural frame. The initial rural NMS frame was based on the 1978 Population Census and later updated with information from the 1988 Population Census. At the beginning, a ward or a group of wards was used as a Primary Sampling Unit (PSU), but later a village was used instead.The rural frame of the NMS was divided into "normal," "large town surroundings" and "low density" strata. In total, 150 strata were created and 2 to 8 PSUs (villages) were selected from each stratum to come up with the sample of villages that can provide estimates for each region of Mainland Tanzania (Module A+B+C). These villages were selected using the probability proportional to size (PPS) selection procedure.The PSUs (villages) for Module A of the rural NMS are automatically included in the regional sample. (b) Urban frame.The urban frame for the NMS was the sample used for the 1988 Population Census detailed questionnaire. For each district in a region, a list of the urban EAs was compiled and a specific number of EAs was selected from this frame using the systematic random sampling (SRS) procedure to produce the regional urban sample.

Sample size, losses and replacement: The final sample analysed for the 2000/01 HBS consisted of 22,178 households, a large sample for any household budget survey.Three PSUs were lost entirely from the sample. Households were included in the analysis if they had at least one record in both the roster and the monthly diary.The weights were calculated for this group of households. Field supervisors were supplied with a list of twelve ‘replacement’ households, drawn as a separate sample at the same time as the main household sample, to be used if a sampled household could not be interviewed for the duration of the survey.The 2000/01 HBS sample had a high level of replacement of households that were not interviewed – around 12 per cent. A total of 4,823 households were analysed for the 1991/92 sample. Losses were higher; levels of replacement were lower (Table A1.2 in main report). In both surveys, households that were part of the initial selection constitute around 85 per cent of the sample analysed.