Socio-Economic Survey 1999

A two stage stratified sampling design with the villages as the first stage units (PSU's) and households as the second stage units(SSU's) was used in the sampling strategy which was based on the method of inter-penetrating sub-samples. A truncated frame which excluded 4.5% of the villages was used because of the difficulty of conducting field work for security reasons in the excluded villages. The survey covered all non-institutional households including one person households. CSES 1999 sampled 6,000 households distributed in 600 villages in the country. The survey was conducted in two rounds to capture seasonal changes in the characteristics studied. The sampling design provided for estimates to be prepared for the urban and rural sectors and the capital city of Phnom Penh as well as for the four ecological zones of the Plain, Tonle Sap Lake, Coastal and Plateau and Mountain Regions. The design is not self-weighting and weights were used in the preparation of survey estimates.

Although CSES 1997 was successful operationally, improvements in the sampling design were considered essential while retaining the main features of the design which has been briefly outlined earlier. The inclusion of such topics as employment, child labor, per capita expenditure of households, health, and education expenditure demonstrated that the sample size should be adequate to produce statistically reliable estimates for the main stratification. The approximate computation of sampling errors of some key estimates in CSES 1997 that sampled 6,000 households showed that the relative errors were in the range of 3 % to 10% or a margin of error twice as much. It was thus necessary to reduce the sampling errors and it was evident that the sampling errors of estimates of the same variables canvassed in CSES 1999 as well as those which had similar prevalence rates would be high. Thus, it was clear that the sample size in fact should be raised above 6,000 households to produce nationally and sectorally representative and statically reliable estimates in respect of some of the key variables in the core questionnaire and in the income and employment module. Because of financial and administrative constraints, it was not feasible to increase the total number of households to be sampled. Therefore it was necessary to resort to the other options available that of improving the precision of the estimates by adopting a more efficient sampling design and attempting to lower the sampling errors.

When compared with the sampling designs that were adopted in surveys conducted earlier in Cambodia, a more efficient and improved sampling strategy was adopted in CSES 1999. The new sampling strategy has provided for estimates for the urban and rural sector, and the capital city of Phnom Penh as well as for the different ecological zones. The method of interpenetrating sub-samples has also provided for the preparation of separate estimates for ecological zones of the country from independent sub-samples enabling checks on the quality of data collected and on the precision of the estimates. Apart from these major innovations, the sampling procedures for the selection of villages which were the primary sampling units (PSU's) and households which formed the secondary sampling units (SSU's) were also improved by adopting circular systematic sampling with probability proportional to size (CSSPPS) techniques.

The sampling design of CSES 1997 followed the sampling strategy adopted in the two socio-economic surveys conducted earlier namely SESC 1993/94 and SESC 1996. The designs in these surveys were based on the division of the country into three domains Phnom Penh, other urban and rural areas so that separate estimates can be prepared for the capital city, and urban and rural sectors. These surveys used truncated frames that had excluded provinces, communes and villages in which data collection was difficult for security or other reasons. From each domain a specified number for villages were selected as first stage units (PSUs) and the second stage units (SSUs) which were households were selected after a pre-listing of households in the sample PSUs.

CSES 1999 sampled 6,000 households from 600 sample villages distributed in all 24 provinces in the country. The survey covered both urban and rural areas of Cambodia. Approximately 4% of the villages were excluded in 14 provinces because of difficulties of conducting field work for security reasons. The number of households sampled from each village was restricted to 10 to reduce the cluster effect and improve the precision of the estimates.

Estimates for round one and round two are provided separately for certain characteristics in addition to the estimates from both rounds of the survey. The estimates provided in the report are for the truncated frame used in the survey that excluded 4.2% of the villages because of the difficulty of conducting fieldwork for security reasons. In respect of a few key variables extrapolated estimates were prepared which covered the excluded areas of the country in addition to the truncated frame used in the survey.