Replicating Replication : Due Diligence in Roodman and Morduch’s Replication of Pitt and Khandker (1998)

Pitt, Mark M. ; Khandker, Shahidur R.

World Bank, Washington, DC

关键词: AGRICULTURE;    ALGEBRA;    BIASES;    CALCULATION;    COEFFICIENTS;   

DOI  :  10.1596/1813-9450-6273
RP-ID  :  WPS6273

学科分类：社会科学、人文和艺术（综合）

来源: World Bank Open Knowledge Repository

【 摘 要 】

"The Impact of Microcredit on thePoor in Bangladesh: Revisiting the Evidence," by DavidRoodman and Jonathan Morduch (2011) is the most recent of asequence of papers and postings that seeks to refute thefindings of the Pitt and Khandker (1998) article "TheImpact of Group-Based Credit on Poor Households inBangladesh: Does the Gender of Participants Matter?"that microcredit for women had significant, favorableeffects on poverty reduction. In this paper the authors showthat these latest Roodman and Morduch claims are based onseriously flawed econometric methods and theory and a lackof due diligence in formulating models and interpretingoutput from packaged software. On the basis of Roodman andMorduch's preferred two-stage least squares regression,an alternative calculation of the standard errors would leadone to conclude that the problem with Pitt and Khandker isthat they underestimate the positive and statisticallysignificant effect of women's credit on householdconsumption. As in their previous efforts, the methods ofRoodman and Morduch are shown to bias the findings in thedirection of rejecting the results of Pitt and Khandker. Wealso further examine two aspects of our instrumentalvariable approach that have been attacked by Roodman andMorduch. The first is the validity of the exclusionrestrictions underlying the use of interactions betweenprogram choice and the set of exogenous variables (includingthe village fixed effects) as instruments. The second is theapplication of the "one-half acre" programeligibility rule. The authors show that identification doesnot require both of these, and present new results droppingeach assumption in turn. The results originally reported inthe Pitt and Khandker paper hold up extremely well in thisnew analysis.