Quantity vs Quality: A Survey Experiment to Improve the Network Scale-up Method

• In Event: 428 - Section on Methodology Paper Session. Creative Methodological Approaches to Data Collection and Analysis

Abstract

Many important problems in science and policy require estimates of the sizes of hidden populations. For example, scientists and policymakers need estimates of the sizes of key populations at risk for HIV – such as sex workers and people who inject drugs – in order to design services, evaluate interventions, and predict the future course of the epidemic. The network scale-up method is a promising technique that uses sampled social network data to estimate the size of hidden populations. We present a theoretical and empirical framework for optimizing a critical but previously unstudied aspect of the method: the trade-off between the quantity and the quality of information collected from respondents. In a large, nationally-representative survey experiment from Rwanda in 2011, we find that collecting less information per respondent, somewhat paradoxically, produces more accurate size estimates for some groups. Our estimates for the sizes of four key populations at risk for HIV in Rwanda are consistently higher than earlier estimates made using other methods in Rwanda, but they are generally lower than international benchmarks. Because our estimates depend on a number of assumptions, we also develop a framework for robustness analysis that enables other researchers to easily produce estimates based on different assumptions.

Authors

• Dennis Feehan, Princeton University

• Aline Umubyeyi, University of Rwanda

• Mary Mahy, Joint United Nations Programme on HIV/AIDS (UNAIDS)

• Wolfgang Hladik, Centers for Disease Control and Prevention

• Matthew J. Salganik, Princeton University