Orphan Survey or Orphaned Survey? Design Effects in a Survey of Orphans and Vulnerable Children

• Activity Number: 68
• Type: Topic Contributed
• Date/Time: Sunday, July 29, 2012 : 4:00 PM to 5:50 PM
• Sponsor: Statistics Without Borders
• Author(s): Bryan Sayer*+
• Companies: Social and Scientific Systems, Inc.
• Keywords: design effects ; Uganda ; orphans

Abstract

In a survey of orphans and vulnerable children (OVCs) in Uganda designed to compare current status of OVCs after the termination of the Track 1 program across partners, the design effects vary substantially by partner and estimate. We examine the variances and design effects in an attempt to determine why they vary so substantially, and differently, by partner. The sample design is stratified by partner, but the ultimate effect is one of clustering. A primary goal of the study is to compare how well partners transitioned OVCs to new aid programs at the termination of the Track 1 program. The majority of estimates are proportions of OVCs receiving different types of aid after the termination of the program. Comparisons of these proportions by partner are hampered by large variances for selected measures. While mean and median design effects are similar for three partners, one partner has design effects of about double the other three, and one partner has design effects (and estimates) near zero. Even among the three partners with similar mean and median design effects, the pattern by estimate varies noticeably.