Questionnaire contents and interviews

Table 2.3 Comparison of NRVA 2003, 2005 and 20078 NRVA 2003 NRVA 2005 NRVA 2007 08 Sampling Sample frame not available; relied on WFP village lists. Expected to be biased for larger rural settlements. Sample frame available from CSO pre- census household listing. Sampling proportional to population, except in the smaller provinces and urban centres where over-sampling insured enumeration of sufficient number of households. Sample frame from updated CSO pre- census household listing. Proportional to population, with over-sampling of smaller provinces and urban centres. Sample selection was not fully random; stratified by different agro-ecological zones. Sample selection based on random selection from geographically ordered PSUs, to give a random spread that represented the spatial distribution of the population. Sample selection based on random start method to have a better geographic distribution of the sample. Household selection was not random within the key issue, but randomly selected within wealth groups. 6-7 households have been assessed within each village. Household selection based on the random start method within villages. 12 households were assessed in each selected village. Household within selected villages was randomly selected from CSO household listing. 8 households in each village selected for assessment. Data collection Four levels of data collection: • District level • Community Shura level • Wealth group level • Household level Three levels of data collection: • District level • Community Shura level • Household level Three levels of data collection: • District level • Community Shura level • Household level Much data collection at the community or district level. Data collection mainly at individual and household level. Data collection mainly at individual and household level. Female interviewers not involved in the south and most eastern areas. Female interviewers participated in all provinces except Zabul. Female interviewers participated in all provinces except Urozgan. Only rural areas and Kuchi population Rural and urban areas and Kuchi population Rural and urban areas and Kuchi population Representing one season, survey conducted within 3 months Representing one season, survey conducted within 3 months Representing all seasons, survey conducted within 12 months Poor female coverage Good female coverage Good female coverage • 32 provinces • 368 districts • 1,853 villages • 5,559 wealth groups • 11,757 rural households • 85,577 persons • 34 provinces • 392 districts • 2,597 clusters • No wealth groups • 30,822 households • 221,586 persons • 34 provinces • 395 districts • 2,572 clusters • No wealth groups • 20,576 households • 152,262 persons Questionnaire contents and design; data processing Partial stakeholder participation in questionnaire design. Full stakeholder participation in questionnaire design. Full stakeholder participation in questionnaire design. Information on basic demographics, health, housing, household assets, migration, risk exposure and response, livestock ownership, agricultural activities and household food consumption. Added information on remittances, HIV AIDS, maternal and child health, household non-food consumption. Added information on disabilities, labour market participation, migration, infant- and under-five mortality, women’s position. More details on household food and non-food consumption. Dropped information on HIVAIDS. Common questionnaire format Teleform scannable questionnaires Teleform scannable questionnaires Some data manually entered; other data transcribed into Teleform format and scanned. Teleform scanning software used for data capture with appropriate Visual Basic on SQL data quality routines built-in. Teleform scanning software used for data capture with appropriate Visual Basic on SQL data quality routines built-in. Methodological backgrounds of NRVA 20078 6 7 Methodological backgrounds of NRVA 20078 NRVA 2003 NRVA 2005 NRVA 2007 08 Staff and training Several trainers involved in training of interviewers, resulting in variation in interviewer performance. Uniform training by 2 trainers involved for the whole country. Uniform training in one training session for all field staff for the whole country. More detail and longer 17 days training. Large number of part-time interviewers. More than 500 field staff were part-time employed. 156 field staff selected. More transparent method of staff selection short listing, test and interview from 12 thousand applicants. Management and funding Managed by WFPVAM Managed by Government CSO-MRRD Managed by Government CSO-MRRD Funded by: • Swiss Government • VAM Headquarters • WFP Funded by: • ECNSS main donor • WFP • UNICEF Funded by: • ECNSS main donor • WFP • DFID • ADB • UNICEF • WB

2.2.5 Data limitations

Even though the lessons learnt during design, implementation and analysis of NRVA 2003 and 2005 were integrated in the design of the NRVA 20078 survey, there are still some constraints and limitations in the data, especially with regard to information on women and children: • Out of the 396 sampled districts, there were five that have not been accessible for security or other reasons in any time of the year. Figure 2.1 provides the accessibility information by fieldwork quarter. • Due to security problems there were no female interviewers in Urozgan province to administer female questionnaire sections. This especially had consequences for the information on reproductive health, child health and fertility and mortality information. Only the food consumption part of female questionnaire sections was collected by male enumerators interviewing male respondents. • After the second month of fieldwork, the worsening security situation prohibited the female interviewers to continue data collection in Nooristan province. Considering the local culture of the province, male enumerators were able to ask the female questionnaire directly to the female respondents. • Analysis of the population structure by sex and age shows under-enumeration of women and girls, and young children, especially infants. Although the coverage of these groups was far better than in 2005, significant numbers are still omitted, a phenomenon that is typical for surveys in many developing countries and especially among Afghan populations. Cultural backgrounds related to the status of women and high infant mortality are likely the reasons for these omissions.