An Evaluation of the Performanceof Regression Discontinuity Design on PROGRESA

Buddelmeyer, Hielke ; Skoufias, Emmanuel

World Bank, Washington, D.C.

关键词: DEVELOPING COUNTRIES;    ESTIMATORS;    ETHICS;    EXPERIMENTAL DESIGN;    EXPERIMENTAL DESIGNS;   

DOI  :  10.1596/1813-9450-3386
RP-ID  :  WPS3386

学科分类：社会科学、人文和艺术（综合）

来源: World Bank Open Knowledge Repository

【 摘 要 】

While providing the most reliable methodof evaluating social programs, randomized experiments indeveloping and developed countries alike are accompanied bypolitical risks and ethical issues that jeopardize thechances of adopting them.In this paper the authors use aunique data set from rural Mexico collected for the purposesof evaluating the impact of the PROGRESA poverty alleviationprogram to examine the performance of a quasi-experimentalestimator, the Regression Discontinuity Design (RDD).Usingas a benchmark the impact estimates based on theexperimental nature of the sample, we examine how estimatesdiffer when we use the RDD as the estimator for evaluatingprogram impact on two key indicators: child schoolattendance and child work.Overall the performance of theRDD was remarkably good.The RDD estimates of programimpact agreed with the experimental estimates in 10 out ofthe 12 possible cases.The two cases in which the RDDmethod failed to reveal any significant program impact onthe school attendance of boys and girls were in the firstyear of the program (round 3).RDD estimates comparable tothe experimental estimates were obtained when we used as acomparison group children from non-eligible households inthe control localities.