Household-based Sample Design Division: Methodology and Evaluation Mr Muthetho Nkwinika 09 June 2008 UNWTO Capacity Building 1
Household Survey Sampling Process Flow UNWTO Capacity Building 2
Population of study The ultimate sampling unit is a dwelling unit where all the households in the selected dwelling unit are enumerated For better control and cost efficiency dwelling units are grouped into enumerator areas (EA) during census 2001 During listing, each dwelling unit within an EA is uniquely numbered and classified according to what the dwelling unit is used for. Only private dwelling units form part of the sample UNWTO Capacity Building 3
Population of study Because of large variation in the sizes of the EAs, Primary Sampling Units (PSU) are formed by combining very small EAs with the medium-sized EAs of similar characteristics and dividing/segmenting very large EAs The measure of size of each EA and formed PSU is the number of households For good precision, the list of PSUs are stratified by province or district council depending on the required reporting level UNWTO Capacity Building 4
Data Exploration and cleaning The list of EAs is explored for correctness and consistency by checking the following: Empty EAs are removed from the frame Small EAs are combined with medium EAs and large EAs are segmented Each EA must be attached to a province, magisterial district, and municipality Each EA must have measure of Size Each stratum must have measure of size UNWTO Capacity Building 5
Precision and cost Precision is the inverse of the variance of survey estimates It determines how good the estimates are The sponsor of the study usually gives either the acceptable precision or the budget Secondary data sources from similar surveys are used to determine the desired precision when it is not given Using a suitable model the sampler balances the precision and the cost to get optimal sample size The objective of balancing the cost and the precision determines the sampling methodology UNWTO Capacity Building 6
Sampling methodology There are several household-based sample surveys conducted per year and as a result, the Master Sample is designed to cater for all of these surveys This means that the Labour Force Survey, the General Household Survey, the Income and Expenditure Survey, the user-paid survey, and ad-hoc surveys are drawn from the Master Sample The number of household-based surveys to be serviced by the Master Sample determines the minimum and maximum measure of size that can be placed on a PSU The sample design follows complex two-stage probability sample The first stage selects PSUs with probability proportional to size (PPS_sys) UNWTO Capacity Building 7
Sampling methodology continues Before the second stage selection can be done, listing of dwelling units is done in the sampled PSUs Listing is to ensure that up-to-date information is available for second stage sampling The second stage selects dwelling units with random systematic sampling Disproportionate or optimal allocation assigns sampling rates to the strata in such a way as to achieve the least variance for the overall mean per unit of cost Using secondary data like previous GHS, the desired overall sample size and design effects are determined UNWTO Capacity Building 8
Errors in Sample Surveys Not all the sampled dwelling units are realized and this result in unit non-response or out-of-scope Survey data suffer from different types of errors including systematic errors, coverage errors, and random errors Non-response occurs in two forms: Unit non-response and Item non-response During sample design, some population elements are erroneously included in the sample resulting in out-ofscope item, e.g. a shop mistaken for a private dwelling During enumeration, some household members are not at home while others refuse to cooperate resulting in unit non-response Respondents refuse to respond to some of the questions which result in item non-response UNWTO Capacity Building 9
Example of a drawn sample UNWTO Capacity Building 10
Conclusion The objective of the study and the type of the target population determine the sample design Sample surveys are expensive to conduct, especially the household based surveys, so it is important that enough effort is put into getting the best out of the process through editing, weight adjustment, and imputation. It is also important to have an idea of the extent of the error in the data to inform decision making process UNWTO Capacity Building 11
Thank you UNWTO Capacity Building 12