On Minimizing the Risk of Bias in Randomized Controlled Trials in Economics

Eble, Alex ; Boone, Peter ; Elbourne, Diana

World Bank, Washington, DC

关键词: randomization;    randomized control trials;    study design;    estimation bias;   

DOI  :  10.1596/1813-9450-7746
RP-ID  :  WPS7746

学科分类：社会科学、人文和艺术（综合）

来源: World Bank Open Knowledge Repository

【 摘 要 】

Estimation of empirical relationships isprone to bias. Economists have carefully studied sources ofbias in structural and quasi-experimental approaches, butthe randomized control trial (RCT) has only begun to receivesuch scrutiny. This paper argues that several lessons frommedicine, derived from analysis of thousands of RCTsestablishing a clear link between certain practices andbiased estimates, can be used to reduce the risk of bias ineconomics RCTs. It identifies the subset of these lessonsapplicable to economics and uses them to assess risk of biasin estimates from economics RCTs published between 2001 and2011. In comparison to medical studies, most economicsstudies examined do not report important details on studydesign necessary to assess risk of bias. Many reportpractices that suggest risk of bias, though this does notnecessarily mean bias resulted. The paper concludes withsuggestions on how to remedy these issues.