March 28, 2011

I just received a heads up from Mark Pitt that he has released a response to Morduch and Roodman's latest critique in the Morduch vs. Pitt & Khandker debate dating back to 1998, on the impact of microfinance in Bangladesh. Before the randomized trials came out Pitt & Khandker's work was the most rigorous attempt to measure the impact of microfinance. Their stunning findings were the centerpiece of evidence for microfinance advocates. But the work has been mired in an intellectual battle over data and methodology for over 10 years. Background here.

In the last round Roodman developed a Stata program to replicate the Pitt & Khandker regressions and Roodman and Morduch find no impact of borrowing. Pitt's new paper again upholds their previous findings and harshly challenges the econometric assumptions of Roodman and Morduch. From his abstract:

Using the dataset constructed by Pitt and Khandker, as well as the data set Roodman and Morduch constructed themselves, the Pitt and Khandker results standup extremely well, indeed are strengthened, when estimated with Roodman’s cmp program, after correcting for the Roodman and Morduch errors.

Pitt makes make some quite technical but obviously serious criticisms of their methodology, and this would appear to be the final word--except that it never is. I know I'm dreaming, but in a case such as this one it would be great if at this point the authors could work together to generate a set of estimators each side agrees is not a glaring econometric error, and then write up separate results reflecting their interpretations. In the meantime we'll keep working on the RCTs...

Pitt's paper is here.