Dean Karlan gave one of three keynote addresses at the recent Randomised Control Trial (RCT) Conference at Monash University, organised by CDES. Here are some points that stood out to me.

We all work in our silos

As researchers and policymakers, we focus on our defined area of expertise. But, poverty traps may have many constraints. Dean pushes for a multidimensional approach and common measures across studies to understand the key constraints. For example, the common measures in the 7 microcredit studies were analysed by Rachel Meager to extract new general insights.

Graduation model

Referring to Science paper where very poor families received several interventions at once.

Comments from Dean:

Graduation Model not a great name. We are not getting people into middle class, just a spectrum on poverty.

Chickens did not work well. High risk - high return. Be careful of the asset you choose.

Pushing against labour model of work vs leisure. Not that simple. Doesn’t capture behaviour.

Cash transfers typically go to consumption and home improvement, and only a small amount to investment.

Six angsts

1. We punish confusing results.
   • creates a publication bias
   • one solution is JDE pre-results review (paper with blank tables). Power is there, papers get rejected because of power.
2. We do not reward replication enough.
   • huge opportunities through pooled data
3. We expect too much of any one study
   • “Catch 22” Fischer & Karlan
4. We do not spend enough time on data quality
   • Andrew Dillon recruited by Northwestern to work on data quality
5. Operations / implementation key
6. Wasted opportunities with one-off projects
   • one solution: long term panel in Ghana (18 hours of survey per household)