Syndicate content

Weekly links June 22: which countries are overrepresented in IEs? How many IEs have data available to replicate them? Mobile savings, and more...

David McKenzie's picture
  • In the Harvard Business Review, Blumenstock, Callen and Ghani summarize their work on using nudges to get government employees to save using mobile money in Afghanistan – “Over six months, the average employee who was enrolled to save by default accumulated an extra half-month’s salary in his or her savings account, relative to employees who had to opt in”
  • An intro to R for Stata users
  • The promise and perils of listening to parents – Sharon Wolf on ongoing efforts in Ghana to improve pre-school quality, and how trying to bring parents onboard backfired.
  • In the Journal of Development Effectiveness, Sabet and Brown track the continued growth of development impact evaluations: “Though we find early evidence of a plateau in the growth rate of development impact evaluations, the number of studies published between January 2010 and September 2015 account for almost two thirds of the total evidence base”. Lots of other interesting facts, including 45% of all impact evaluations occurred in just 10 countries, with Kenya and Uganda having the most impact evaluations per million population, and Sub-Saharan Africa the most commonly represented region – perhaps something for donors to think about...
  • 3ie now has a working paper out on their push-button replication project: they take 109 development IE papers published in 5 public health journals and 5 economics journals and see whether they could run code that produces the main tables and figures. This was possible or possible with minor differences for only 46/109 – the biggest problem was not getting files at all in many cases – particularly for public health journals, where not one paper had publicly available replication data. On a positive note, World Bank funded studies had greater availability of replication data than studies funded by the four other major funders they consider. (I was also pleased to read until the end and find myself singled out for having 5 studies on the list, all of which had publicly available data and code, and all of which replicated.)
  • Karthik Muralidharan makes an argument for skipping the baseline in randomized impact evaluations implemented with governments: "The biggest risk is implementation failure. You’ll spend a lot of time doing the baseline – spend time, spend money – and have the intervention not be implemented. So when you’re working with the government, it’s better to get power by doubling the sample of your endline and just randomized with administrative data so you’ll get the same amount of power but you reduce risk up front." A few commentators push back, reminding of the value of baselines.
  • NEUDC submissions now open, due by July 31.

Add new comment