The Challenge of Sample Design During Civil War

1. The Challenge of Sample Design During Civil War The Case of Southern Sudan Luka B. DENG New Sudan Centre for Statistics and Evaluation Cancun, Mexico, 2004

2. Presentation Outline 1. Introduction 2. Data Needs During Civil War 3. Challenges of Sample Design 4. Traditional Social Hierarchy 5. Conclusion

3. 1. Introduction: Violent Conflict • Civil wars are pronounced and endemic after the Cold War • Highest Incidence of Civil Wars in Africa • Half of African countries affected by civil war • One in three African people affected by civil war • What is “normal” is violence rather than peace in most parts of Africa

4. Violent Conflict in Africa

5. 2. Data Needs During Civil War • Increased incidence of poverty • Increased vulnerability and food insecurity (40% of people at risk of hunger and malnutrition from zones of active conflict in the early 2000) • Recurrent famines (7 out 8 famines caused by civil wars in the 1990s) • Increased humanitarian interventions (relief aid is replacing development assistance in the war affected countries) • Increased need to monitor rural livelihoods in zones of active conflict

6. 3. Challenges of Sample Design • No census tradition or framework for comprehensive survey work • Lack of relevant enumeration areas or sample frame • Erratic and recurrent population movement • Changing administrative boundaries • Lack of cartographical and mapping references • Statisticians dealing only with “normalcy” • Then, how can we make sample design in such a situation of civil war!

7. 4. Traditional Social Hierarchy • Traditional social structure survives in civil wars more than any other institution. • It is clearly defined down to a household level • It is a custodian of community’s knowledge and experiences • It is the basis for any administrative structure during civil war • Use social structure to provide a workable frame for sample selection, although it is not ideal

8. The Case of South Sudan • Recurrent Civil wars (1955-1972, 1982-now) • Agriculture as mainstay of the rural livelihoods • More than 98% rural population • Recurrent famines (1988, 1991, 1993, 1998) • Last population census in 1953 during colonial period, other censuses (1973, 1983, 1993) partially conducted because of civil wars • No agriculture census ever conducted in South Sudan, only in 1965 in Northern Sudan • Data gathering during civil war focuses on livelihood and food security monitoring

9. Overview of Sample Design • Use traditional structure (chiefs, sub-chiefs, headmen and households) as sample frame/enumeration units • Three-stage probability sampling • First stage involves listing of executive chiefs as primary sampling units (PSUs). • Second stage involves selection of sample headmen as second-stage units (SSUs) through systematic sampling from PSUs • Third stage involves random selection of sample households for interview and data gathering

10. Pros and Cons of such Sample Design Advantages: • Probability sample has been used • Control of sample size that limits non-response • Equal and constant interviewer workload • Cost effectiveness by using existing traditional rosters of households rather than a fresh and costly listing by strangers • Sample large enough (5000 HHs) to provide disaggregated estimates

11. Pros and Cons Cont. Disadvantages: • Rather unorthodox and inferior sample method • Less reliable because of sampling error cannot be controlled as a result of lack of reliable size measures for stratification and selection process (population under each chief has been used as a proxy for size measures) • Need to reflect differential probabilities of selection by applying weighting as estimates cannot be prepared from sample counts alone

12. 5. Conclusion • Probability sampling and “implicit stratification” have been in a way followed • Applied in various surveys (Multiple indicators Cluster, Livelihood and Food Security Monitoring and sentinel sites) • Generated satisfactory results • Most appropriate in war situations