National Risk and Vulnerability Assessment Survey 2007-2008

Sampling Design

The sampling design of the NRVA 2007/8 built upon the experiences from the NRVA 2005 and the same sampling approach was maintained, with representative statistics generated for 34 provinces, 11 instead of 10 urban centres, and the nomadic pastoralists. Based on the CSO pre-census household listing data, a geographically ordered list of all of the primary sampling units (PSUs) (rural settlements and urban blocks) was created with their estimated number of households. The resulting sample is a total of 2,441 PSUs and 20,576 households (19,528 from the settled urban and rural population and 131 (1,048 households) from the Kuchi population.

The use of a random start and a pre-determined sampling interval provided each household with an equal chance of being selected. Once individual households had been selected, seven other households were selected within the same PSU. Consequently, in every PSU, eight households were interviewed. To avoid the problem of household replacements in the field, the 2007/8 sample design was improved by randomly selecting four households that were kept as reserve in case the household had moved away.

A major difference with the 2005 assessment is the removal of the seasonality bias by conducting the survey yearround during all 12 months. The main considerations for moving from data collection in summer only to the year-round strategy were:
• to capture the seasonality of consumption and other dimensions of welfare;
• to improve the quality of the collected data;
• to field a smaller group of carefully selected, well trained and properly supervised interviewers;
• to permit higher quality field and data validation procedures, through avoiding a concentration of the sample into just two months;
• to be consistent with international standards for obtaining seasonally biased estimates.

Note: See detailed sampling methodology in Annex II of the final report which is presented in this documentation.