Weekly links December 10: where ideas came from, does the gender of your co-workers matter, pay-for-performance in Nigerian health, and more…


This page in:

·       On VoxDev, Eeshani Kandpal and co-authors summarize their study comparing a pay-for-performance program in Nigeria’s health sector to a simpler intervention that decentralized the financing of health facilities. They find this simpler intervention achieves similar results, at half the cost.

·       I always enjoy the IMF’s Finance and Development magazine’s interviews with economists. The winter 2021 issue has an interview with Amy Finkelstein: I liked this story about how she found out about the Oregon Medicaid lottery “Finkelstein says she keeps a mental list of questions that interest her and an eye out for settings that will help her find the answers. That is what happened in 2008, when the host of a TV comedy show she was watching joked about the state of Oregon’s decision to use a lottery to choose a limited number of people to be enrolled in Medicaid, the health insurance program for low-income adults….“Oh my God, an RCT!” Finkelstein recalls thinking. “We’ve got to get the data!”. The fall 2021 issues interviews Solomon Hsiang about his work on climate change, with discussion of two of the main critiques of his work, and of the type of interdisciplinary team needed for work on this problem.

·       On the IGC blog, Deepshikha Batheja looks at whether the gender of co-workers matters for call center employees via an RCT in five Indian cities. There is zero effect on male worker productivity, some increases in knowledge sharing for male employees, and male employees assigned to mixed gender teams are more likely to be dating. Female employees see no change in productivity, get no knowledge spillovers, but get more peer monitoring and support relative to those in all female teams.

·       A new survey paper by Clément de Chaisemartin and Xavier D'Haultfœuille review the rapidly growing new difference-in-differences and two-way fixed effects literatures with treatment effect heterogeneity – and the R and Stata codes needed to implement these new methods. One point is that the issue of negative weights is more likely for non-binary treatments. There is also a nice little discussion of things to consider when deciding which of the new methods to employ, and that there might be a bias-variance trade-off if parallel trends do not exactly hold, and where there are still some open questions in the literature.

·       The Nobel lectures by Card, Angrist and Imbens are now up on YouTube and good viewing. Quite a difference in styles of the different talks. Card gives a nice intellectual history of where the design approach to identifying causal effects came from in economics, with some interesting notes along the way of how he came up with some of his most famous papers – for example, that it was an undergrad student who told him about the Mariel boatlift; Angrist takes advantage of the virtual format for a slick and lively video production, with lots of animated videos to explain his work on education, the theory of LATE; the importance of being clear on what the counterfactual to getting admitted into a treatment school really is; and a fun video of what the computers he worked on early in his career looked like;  from Imbens talk, 50% of NBER papers now use the term causal; he gets Dad of the year by quoting his 10-year old daughter in his Nobel lecture; has a great timeline and history of causality work; the concern about defiers in judge leniency designs; a defense of LATE as being an estimate of interest; and more…


David McKenzie

Lead Economist, Development Research Group, World Bank

Join the Conversation