Weekly links Feb 1: g big data, scaling up CCTs, “the data have been mined, of course”, and more...


This page in:

  • Working with big datasets in Stata? Then the package gtools might be for you – I love that they have to give the caveat “Due to a Stata bug, gtools cannot support more than 2^31-1 (2.1 billion) observations”. Meanwhile, the Stata blog has the second post on doing power calculations via simulations in Stata.
  • More on industrial policy: A nice summary at VoxDev by Ernest Liu of his work on industrial policies in networks, and a reason to prioritize upstream sectors.
  • New SIEF note on using phone monitoring to help more money reach target beneficiaries: an example where small effects are meaningful when cheap and scaled to many people – the treatment group were only 1.3% more likely to get their money, but this meant about $1 million more funding reached farmers when officials knew they would be phone monitored, and the monitoring only cost $36,000.
  • Youtube video of Rohini Pande’s Carolyn Shaw Bell award, featuring a lovely tribute by Natalia Rigol which highlights the many, many ways Rohini has helped the profession, and insightful thoughts from Rohini on what the profession needs to do better.
  • The latest issue of the JEP also has a symposium on gender in economics with suggestions of what can be done to better attract, support, and retain women in the economics profession, including Kasey Buckles’ paper summarizing what impact evaluations have shown to work in this regard.
  • Scaling up CCT pilots – Apolitical talks with Damien de Walque about how his work on CCTs in Burkina Faso was able to be scaled up because in part of the government using it as a training ground from the start (h/t Rachael Strohm)
  • A glimpse into past practice: Jishnu Das shared with me this description by the authors (Summers and Wolfe) of a 1977 AER paper about their empirical work:“In winnowing down the original list of variables to get the equation of "best fit," many regressions have been run. A variety of functional forms non-linear, interactive for the variables have been examined. The data have been mined, of course... We threaded our way through the myriad of variable combinations with the magic 5 percent normal curve numbers as our North Star. But, to the final formulations and interpretations of coefficients, more informal standards were applied... Variables which had coefficients whose significance were very sensitive to the introduction and discarding of other variables were not retained” - this is definitely a certain form of transparency.
  • Job openings: EDI (a subsidiary of Mathematica) has a senior research officer and assistant research officer position, both for the running of large-scale surveys and impact evaluations (on health, agriculture, education, infrastructure). 


David McKenzie

Lead Economist, Development Research Group, World Bank

Join the Conversation

February 01, 2019

The link to the Rohini talk actually leads to the Mellissa Dell talk. Here is the Rohini one: