Weekly links July 17: Regret and behavior change, DAGs vs potential outcomes, a requiem for Charter Cities, and more...

|

This page in:

·       This Freakonomics episode on Behavior Change features an all-star cast including Angela Duckworth, Katy Milkman and Richard Thaler talking about some of their latest work and findings on behavioral science – the whole episode is interesting and a good example of communicating results and experiments to a general audience. Tom Gilovich discusses research on regret, and whether you regret action or inaction more – something to ponder as you decide whether to click on the link.

·       How IDinsight uses Stata, R, and Python for different needs

·       On the Devpolicy blog, Matthew Woolf discusses the many reasons why charter cities has failed.

·       Following up on last week’s links discussion on third-party replicability, this week the AEA’s updated its data availability policy for papers published in AEA journals,  and the AEA’s new data editor, Lars Vilhuber, set up a twitter account (@AeaData)  in which he said they will now attempt to run people’s code, will try to have third party replication of restricted access data sets, and gave a template for writing a better readme file.  It will be interesting to see how this works in particular with some of the structural models that take weeks to run the code, and with studies that use tax data and other restricted access data.

·       A new paper by Guido Imbens discusses the Directed Acyclic Graph (DAG) approach to causality, compares it to the potential outcomes framework more commonly used in economics, and discusses pros and cons of each method, and why DAGs have not caught on much in economics.  This includes a review of the Book of Why. This is very nicely written, and a great comparison of the two approaches to thinking about identifying causal effects.  His verdict: “I think the case for the DAGs is the strongest in terms of exposition. The DAGs are often clear and accessible ways to expressing visually some, though not necessarily all, of the key assumptions...” but DAGs are not so useful at capturing shape restrictions like monotonicity, treatment effect heterogeneity, equilibrium behavior, and several other phenomena important in economic estimation, leading him to conclude that “The questions it currently answers well are not the ones that are the most pressing ones in practice.... for the problems where DAGs could contribute substantially, the most important issue holding back the DAGs is the lack of convincing empirical applications... more papers discussing toy models will not be sufficient.”

·       The Atlantic has a long piece on Raj Chetty, discussing his upbringing in India, the role of chance in shaping his life, the work of his new Opportunity Insights initiative, and much more. It also describes how he starting work with U.S. tax data: “In November 2007, Chetty came across an ad from the IRS seeking help organizing its electronic files into a format that would be easier to use for research. He immediately recognized that completing the job would make it possible for scholars to go far deeper into tax data. He and John Friedman began the process of registering to be federal contractors—which involved, among other things, certifying that their workplace met federal safety standards, and calling on Friedman’s brother, who lived in Washington, D.C., to take a cab out to Maryland to hand-deliver their application materials, in triplicate. Like many good ideas, the project seems obvious in retrospect, but the truth is that nobody could have known how useful the data would prove to be

Authors

David McKenzie

Lead Economist, Development Research Group, World Bank

Join the Conversation