Published on Development Impact

December 11 links: Expensive publishing, DAGs & linking errors, waning research on China and where is overstudied, and more…

This page in:

·       A review paper in Science on the links between poverty, depression and anxiety by Ridley, Rao, Schilbach and Patel – including a round-up of experimental impacts of poverty reduction projects on mental health, and of mental health improvement interventions on labor supply.

·       In the Royal Economic Society newsletter, Dina Pomeranz offers an introduction to #Econtwitter

·       The Economist on how research interest on China is waning as its influence rises.  ICYMI, here is my blog from earlier in the year on doing RCTs or fieldwork in China.

·       At VoxDev, Suresh de Mel and co-authors summarize their experiment in Sri Lanka that allowed people to make deposits to their savings account from their mobile phone – which found very little use and no impact on savings.  Documenting these null results is particularly important for areas such as the digital economy and mobile money which get so much hype.

·       In the Economist, an analysis of which countries research is done on, from 900,000 papers published 1990-2019 – finds a very strong correlation between the overall size of the economy and number of papers – “The 70 least-studied countries account for just 1% of all mentions in economics papers over the past three decades”. Other key correlates were data availability, having English as an official language, and sending lots of students to U.S. universities.

·       JPAL’s guide for ethical conduct in randomized experiments

·       Is this the first time that twitter discussion of econometric methods has made it into the JEL? Guido Imben’s paper comparing the potential outcome and directed acyclic graph (DAG) approaches to causality contains several quotes directly from twitter discussing pros and cons of these methods – as well as a great summary of why DAGs have not caught on in economics.

·       Also in the JEL, Bailey et al. discuss the use of automated linking methods (e.g. using different records like birth certificates, historic censuses, etc. and  trying to link individuals by name, gender, race, place of birth and other such characteristics) and compare them to ground-truth – finding that measurement error can be large and systematically correlated with sample characteristics. This is important for the use of big data approaches that rely on automated linking of records, and they give recommendations for ways to improve.

·       A useful short paper by Campbell Harvey and David Hirshleifer on improving peer-reviewed publishing. One of their points is to push editors away from a “Union heuristic” of requiring authors to perform all checks and extensions suggested by referees; and more push towards Up or Out decisions. They also call for more experimentation among journals on different models, and note the example of the Review of Finance, which “In addition to the normal review process, offers a special review process to authors seeking a very fast editorial decision at a considerably higher submission fee (900 euros!). This “Fast-Track” process guarantees an editorial decision in 14 days” – watch out though if that decision is an R&R – you then can pay 500 euros for a resubmitted Fast Track! Still a bargain compared to the 9,500 euros that Nature will charge to make your paper open access

·       Devpolicy on a new Pacific Data Hub

·       Andrew Gelman and Aki Vehtari on the 8 most important statistical ideas of the last 50 years – causal inference, the bootstrap, robust inference, and overparameterized models with regularization (such as many machine-learning approaches) are among them – and also interesting to see ideas which have had less traction/use in economics.



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