Working Paper: NBER ID: w30762
Authors: Sylvain Chassang; Samuel Kapon
Abstract: This paper describes a number of strategies that experimenters may use to improve the external validity of their own findings, and of their research field as a whole. The paper emphasizes a dynamic view of research processes, in which learning about treatment and treatment adoption does not cease after a given study is performed. External validity need not be an unattainable goal in such a context. However, because researchers today need not be the same as researchers and policy-makers tomorrow, dynamic research processes are affected by research externalities, i.e. research practices that have high social value but low private returns. The paper identifies several of these research externalities and argues that funding organizations can have a significant impact at a relatively modest cost by subsidizing external-validity add-on modules specifically targeting research externalities.
Keywords: external validity; randomized controlled trials; research practices; funding organizations
JEL Codes: C90; C93
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| improving external validity (C90) | positive externalities from research practices (O36) |
| funding organizations subsidizing addon modules (L17) | improved external validity (C90) |
| internal validity (C90) | enhancing external validity (C90) |
| research practices that prevent biases (C90) | internal validity (C90) |
| dynamic understanding of research processes (C90) | treatment adoption adjusts to new evidence (C22) |
| rich covariates and diverse participant contexts (C90) | strong assumptions about treatment effects hold (C22) |