On the use of demographic and survey estimation techniques to estimate enforced disappearances: a case-study of Punjab, India

Romesh Silva, University of California, Berkeley
Patrick Ball, Human Rights Data Analysis Group, Benetech
Jeff Klingner, Human Rights Data Analysis Group, Benetech

When evaluating the magnitude and pattern of conflict-related mortality during armed conflicts, demographers rely on either available population census data, convenience sample data, retrospective mortality data or a combination of these data sources. By its nature, each source is vulnerable to bias and error. Existing quantitative claims about enforced disappearances in Punjab (India) between 1984 and 1996 are not based on defensible demographic methods. We review and compare the applicability of Multiple Systems Estimation (MSE) and quasi-adaptive survey sampling designs for measuring the magnitude of these enforced disappearances. By drawing on six independently collected data sources, we derive MSE estimates of the magnitude and pattern of total enforced disappearances in Punjab. These estimates are based on probabilistic record-linkage methods which draw heavily from recent advances in statistical learning. We also present estimation results from a quasi-adaptive sample survey of families of the disappeared in Punjab.