Published on Development Impact

Weekly links August 1, 2019: summer learning resources, using baseline surveys, when to cross-cut, and more before we break for August

This page in:

·       The latest SIEF newsletter offers multiple examples of how teams have been using baseline surveys for research – more nice examples to add to those in our posts on Getting more than one paper out of a given impact evaluation and publishing descriptive papers, and Dave Evans’ 5 uses for a baseline survey (h/t Noel Muller).

·       Max Kasy’s lecture slides for his course on advances in causality and foundations of machine learning. He also points to this forthcoming JEL paper on “text as data” by Gentzkow et al. that I hadn’t seen.

·       Lecture videos from Oxford CSAE’s meta-analysis in development economics workshop.

·       Following up on my post this week on cross-cutting designs, Macartan Humphreys reminded me of the guest post we featured from the Declare Design team he is part of, which also discussed the issue of trade-offs between factorial (2x2) designs and three-arm designs (see also their recent paper). They note that researchers often choose 2x2 designs because they expect interaction effects, but, perhaps counterintuitively, the designs may actually be best when there are no (or very weak interactions) – in which case using them instead of three-arm trials and ignoring interactions can boost power with little bias. Of course the problem is that in practice it is hard to know whether the interactions are there or not, given often low power to detect interactions.

·       Funding call for proposals: The IZA/DFID program on gender, growth and labor markets in low-income countries has its first call for proposals open, with both large and small research grants available.

Development Impact will take a break for August. See you after Labor Day.


David McKenzie

Lead Economist, Development Research Group, World Bank

Join the Conversation

The content of this field is kept private and will not be shown publicly
Remaining characters: 1000