#randomized control trials

[daisukitoo]

Found in the wild: explicit use of the Copenhagen interpretation of ethics, down to the use of the term “entangled.”

I am unclear on whether Professor Hellman is advocating that position or just saying that it exists. Given the title of the piece, “An Experiment Gives Cash Aid To The Poor. Is That Ethical?“ NPR seems to be taking the Copenhagen side, especially if that title is intentionally invoking Betteridge’s Law.

[tofutiger]

Randomization in the Tropics Revisited

[nobenchmark-blog]

Lean Development

                The Lean Startup and Poor Economics may seem like very different boks. While the first is a business book for startups, the second one is about development economics. Yet after reading them back to back, I realized they are not only about the same thing (learning), but that together they suggest a very slick method to think about economic development. 

In Poor Economics, Esther Duflo and Abhijit Banerjee (D&B) argue that “so much of anti-poverty policy has failed over the years because of an inadequate understanding of poverty.” They then go on to share how hundreds of rigorous randomized controlled trials have helped us learn about poverty, and explain how using that knowledge is key in the fight against it.

In The Lean Startup, Eric Ries tells us that startups fail because they have an inadequate understanding of who their customers are, or even of what their product should be. He then goes on to argue that in that uncertain environment startups exist to learn, and that the learning process holds the key to a startup’s success.

Both books are excellent reads alone, but together they were a real source of inspiration. Combining rigorous testing methods to learn about poverty with practical startup management methods could help development practitioners reach some transformational outcomes. Imagine how much great stuff could come out of a crossover between the World Bank and Dropbox!

But how would that actually work? One way would be to apply “lean methods” to development projects such as the build-measure-learn loop as described by Ries:

Build – Create a minimum viable product (i.e. enough of a product to allow you to learn)

Measure – Test your assumptions against an appropriate accounting system Learn – Decide how to proceed or change course

But are randomized control trials anything other than a way to build, measure and learn from evidence? I think they are pretty close.

There are two lean startup concepts, however, which are not obvious in D&B’s work and which I think we could stress more in development today.  

The first is the use of a minimum viable product (MVP), which is the idea that early products should focus on collecting the maximum amount of learning from a minimum amount of effort. A minimum viable development project (MVDP) would be a ruthlessly cheap development experiment that would allow us to learn about development outcomes D&B-style. For example, simple mobile classrooms or clinics could test a community’s needs before the decision to build a full-fledged school or hospital. And while I acknowledge that pilot projects are not something new, and are the basis of much of the learning in Poor Economics, the emphasis here is on the words “maximum” and “minimum.”

The second is the idea of acceleration, which is the quick iteration through the build-measure-learn cycle. Learn quickly. Decide quickly. Avoid waste. While institutions like the World Bank are indeed focused on learning and their project cycle does include evaluation, the timeframe for most projects is very long. And while in some instances development outcomes do require a much longer timeframe of evaluation, I also think there are projects out there that could be launched, tested and changed in a matter of days, creating a great deal of learning and improving development effectiveness.

I acknowledge that policy-makers often operate under tighter constraints (including political ones) and have to solve for many more variables than business leaders, but still there should be some startup lessons learned that we can apply to policy-making and development.