Multiple Indicator Cluster Survey 2010

The primary objective of the sample design for the Suriname Multiple Indicator Cluster Survey was to produce statistically reliable estimates of most indicators, at the national level, for areas classified as urban, rural coastal and rural interior, and for the 10 districts of the country - Paramaribo, Wanica, Nickerie, Coronie, Marowijne, Commewijne, Sarramacca, Para, Brokopondo, and Sipaliwini.

A multi-stage, stratified cluster sampling approach was used for the selection of the survey sample.

The target sample size for the Suriname MICS was calculated as 9,000 households. For the calculation of the sample size, the key indicator used was the number of children younger than 5 years of age who had had diarrhoea in the past two weeks before the survey using the estimate of the last MICS3 survey.

Suriname is divided into 10 districts and 62 'ressorten' by law. The 'ressorten' are subdivisions at the district level. For purposes of conducting the fieldwork during the Seventh Population and Housing Census the General Bureau of Statistics subdivided each ressort in the coastal area (lowland and savannah) into 'telblokken'. A 'telblok' also called an enumeration block, was considered to be the manageable workload for a Census enumerator for the fieldwork period of two weeks and would ideally have between 100 and 150 objects. An object can be any kind of building or a construction work, like, churches, schools, stores, houses, dwellings etc. In order to clarify: not every object stands for a dwelling or living quarters of a household. In the interior (rainforest) a somewhat different fieldwork approach was used, whereby teams consisting of 5-7 fieldworkers canvassed clusters of villages. These clusters were called 'telgebieden' and were expected to have approximately 500 households, or the workload of 5 interviewers. "Telgebied" can also be called an enumeration area.

The 2004 census frame was used for the selection of clusters as the results of the 2004 Census provide a basis for provisional estimates on the number of households. Thus, the 'telblokken' and 'telgebieden' were considered the best currently available subdivisions by the General Bureau of Statistics and formed the basis for the MICS 2010 sample design.

Each of the 10 districts in the country is allocated to one of these strata, but with three towns (Nw. Nickerie in Nickerie district, and Meerzorg and Tamanredjo in Commewijne district) being counted as urban, even though they are located in what are otherwise rural districts.

In the case of MICS3, the total sample had been about 6,000 households. Survey results had been reported not only for the three strata, but also for a five-way breakdown of districts. This was achieved by grouping districts as follows: Paramaribo; Wanica and Para; Nickerie, Coronie and Saramacca; Commewijne and Marowijne; and Brokopondo and Sipaliwini. In the case of MICS4, the sample size was increased to 9,000 households. One of the main benefits of this increase was that it would permit the reporting of indicators at the district level.

The allocation within each stratum was done with probability proportional to size, where the population of each area (from the 2004 census) was used as the measure of size. It was only after the fieldwork was completed that it was realized that the samples allocated to several of the individual districts were insufficient to provide satisfactory estimates for many of the variables. A more equal allocation to each district would have provided more precise estimates for the smaller districts.

It should be noted that, according to sampling theory, it is the size of the sample, rather than the proportion of the population covered, that is the key factor in determining the precision of the estimate. Several districts have sample sizes that are around the 500 level, which is slightly on the low side. An allocation of about 700 households would have been more appropriate, which could have been achieved by reducing the allocation for Paramaribo. In the rural interior, the allocation for Brokopondo might usefully have been increased, with a corresponding reduction in Sipaliwini. Only 140 households were allocated to Cornie, reflecting its small population of less than 1,000 households, but it would have been necessary to cover a larger number of households there (say 300 or 400) in order to obtain reliable estimates.

The actual sample selection in the selected clusters was done as follows. In urban and rural coastal areas, where enumeration districts (EDs) usually contain about 150 households, one pointer address (PA) was selected at random within the ED. If it was not the address of a private household, the next address was taken as the starting point. Twenty adjacent addresses (1 to 20) were then selected around this PA, and a printed map provided to each team, showing the location of each address. In rural areas the enumeration areas might consist of either one village or several smaller villages combined. Where a village was very isolated, it was treated as one enumeration area, even though sometimes it did not contain many households.

The sampling procedures are more fully described in "Suriname Multiple Indicator Cluster Survey 2010 - Final Report" pp.178-181.